cs user al simulations - gromacs

134
GROMACS USER MANUAL Groningen Machine for Chemical Simulations Version 3.0

Upload: others

Post on 18-Oct-2021

21 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: CS USER AL Simulations - GROMACS

GR

OM

AC

SU

SE

RM

AN

UA

LG

ron

inge

nM

ach

ine

for

Ch

em

ica

lSim

ula

tion

s

Vers

ion

3.0

Page 2: CS USER AL Simulations - GROMACS

Don’tforgetto

checkoutthe

onlineresources

atw

ww

.gromacs.org.

Page 3: CS USER AL Simulations - GROMACS

24

6In

dex

tpbc

onv

226

traj

ecto

ryfil

e26

,119

tree

38tr

jcat

226

trjc

onv

227

trjo

rder

228

type

atom

∼se

eato

mty

peF

ile∼

seeF

ilety

pe

U umbr

ella

sam

plin

g10

3un

ited

atom

81

V velo

city

,cen

ter-

of-m

ass

seec

ente

r-of

-mas

sve

loci

tyvi

brat

ion,

angl

ese

eang

levi

brat

ion

viria

l20

,70,

71,1

57vi

rtua

lsite

81V

isco

sity

111

visc

osity

129,

140

W wat

er56

wea

kco

uplin

g22

,24

whe

el14

8,22

9

X x2to

p22

9xd

r11

5xm

drun

230

xmgr

138,

175

xpm

2ps

231

xram

a14

8,23

2

GR

OM

AC

SU

SE

RM

AN

UA

LVe

rsio

n3.

0

Dav

idva

nde

rS

poel

Ald

ertR

.van

Buu

ren

Em

ileA

pol

Pie

ter

J.M

eule

nhof

fD

.Pet

erT

iele

man

Alfo

nsL.

T.M

.Sijb

ers

Ber

kH

ess

K.A

nton

Fee

nstr

aE

rikLi

ndah

lR

udiv

anD

rune

nH

erm

anJ.

C.B

eren

dsen

Cop

yrig

ht(C

)19

91–2

001

Dep

artm

ento

fBio

phys

ical

Che

mis

try,

Uni

vers

ityof

Gro

ning

en.

Nije

nbor

gh4,

9747

AG

Gro

ning

en,T

heN

ethe

rland

s.

Page 4: CS USER AL Simulations - GROMACS

ivPreface

&D

isclaimer

This

manualis

notcomplete

andhas

nopretention

tobe

sodue

tolack

oftime

ofthecontributors

–our

firstpriorityis

toim

provethe

software.

Itism

eantasa

sourceofinform

ationand

referencesforthe

GR

OM

AC

Suser.

Itcoversboth

thephysicalbackground

ofMD

simulations

ingeneraland

detailsof

theG

RO

MA

CS

software

inparticular.

The

manualis

continuouslybeing

worked

on,w

hichin

some

casesm

ightmean

theinform

ationis

notentirelycorrect.

When

citingthis

documentin

anyscientific

publicationplease

referto

itas:

vander

Spoel,D

.,A.R

.vanB

uuren,E.A

pol,P.J.Meulenhoff,D

.P.Tielem

an,A

.L.T.M.S

ijbers,B.H

ess,K.A

.Feenstra,E

.Lindahl,R.van

Drunen

andH

.J.C.

Berendsen,G

rom

acs

Use

rM

an

ua

lversio

n3

.0,

Nijenborgh

4,9747A

GG

roningen,The

Netherlands.

Internet:w

ww

.gromacs.org(2001)

or,ifyouuse

BibTeX

,youcan

directlycopy

thefollow

ing:

@M

an

ua

l{gm

x30

,title

="G

rom

acs

{U}se

r{M

}an

ua

lve

rsion

3.0

",a

uth

or

="D

avid

van

de

rS

po

el

an

dA

lde

rtR

.va

nB

uu

ren

an

dE

mile

Ap

ol

an

dP

iete

rJ.

Me

ule

n\-h

off

an

dD

.P

ete

rT

iele

ma

na

nd

Alfo

ns

L.

T.

M.

Sij\-b

ers

an

dB

erk

He

ssa

nd

K.

An

ton

Fe

en

straa

nd

Erik

Lin

da

hl

an

dR

ud

iva

nD

run

en

an

dH

erm

an

J.C

.B

ere

nd

sen

",a

dd

ress=

"Nij\-e

nb

org

h4

,9

74

7A

GG

ron

ing

en

,T

he

Ne

the

rlan

ds.

Inte

rne

t:h

ttp://w

ww

.gro

ma

cs.org

",ye

ar

="2

00

1"

}We

humbly

askthat

youcite

theG

RO

MA

CS

papers[

1,2]

when

youpublish

yourresults.

Any

futuredevelopem

entdependson

academic

researchgrants,since

thepackage

isdistributed

asfree

software!

Any

comm

entsare

welcom

e,pleasesend

themby

e-mailtogrom

acs@grom

acs.org.

Groningen,A

ugust10,2001.

Departm

entofBiophysicalC

hemistry

University

ofGroningen

Nijenborgh

49747

AG

Groningen

The

Netherlands

Fax:

+31-503634800

Ind

ex2

45

polarizability28

polymer

convention84

Position

restraint61

positionrestraints

117potential

energy20

function47,113

potentialsofm

eanforce

103P

PP

M44,75,121

precisiondouble∼

153single∼

153pressure

20P

ressurecoupling

23,123P

arrinello-Rahm

an∼se

eParrinello-R

ahman

pressurecoupling

principalcomponentanalysis

seecovariance

analysisprocessor

topology38

Program

sby

topic129

Proper

dihedral5984

protonate225

pulling,AF

Mse

eAF

Mpulling

QQS

AR

1quadrupole

81

RReaction

Field

67reaction

field50,167

Reaction-F

ield121

refinement,nm

r62

repulsion48

restraintA

ngle∼

seeA

nglerestraint

Distance∼

seeD

istancerestraint

Position∼

seeP

ositionrestraint

rule,combination

seecom

binationrule

Run

Param

eter117

Ryckaert-B

ellemans

84

Ssampling

26um

brella∼se

eumbrella

sampling

Schrodinger

equation1

searchG

rid∼

seeG

ridsearch

Sim

ple∼se

eSim

plesearch

searching,Neighbor

seeN

eighborsearching

SE

TT

LE29,86

SH

AK

E29,68

shake126

sharedm

emory

45shell

seeparticle

model

28S

hellMolecular

Dynam

ics119

shiftfunction19

Sim

plesearch

18S

imulated

annealing12531

singleprecision

seeprecision,single

Soft-core

interactions68

solver,Poisson

seeP

oissonsolver

Space

Decom

position38

statisticalmechanics

2S

teepestDescent

33steepestdescent

117S

tochasticD

ynamics

32stochastic

dynamics

2stretching,bond

seebond

stretchingS

urfacetension

coupling25

TTabulatedfunction

165tem

perature20

Temperature

coupling21,123

temperature

coupling14,21

Berendsen∼

seeB

erendsentem

peraturecoupling

temperature

coupling,Nose-H

ooverse

eNose-H

oovertem

peraturecoupling

terminidatabase

89third

neighbor69

time

lag138

timestep,integration

seeintegration

timestep

topic,Program

sby

seeP

rograms

bytopic

topology79

Topologyfile

91topology,processorse

eprocessortopology

Page 5: CS USER AL Simulations - GROMACS

24

4In

dex

inst

all

153

inte

grat

ion

times

tep

56in

tera

ctio

nlis

t69

1-4∼

see1

-4in

tera

ctio

nis

othe

rmal

com

pres

sibi

lity

24

K kine

ticen

ergy

20

L Lang

evin

dyna

mic

s32

,118

leap

-fro

g21

,117

Lenn

ard-

Jone

s48

,67

limita

tions

3LI

NC

S29

,68

lincs

126

list,

inte

ract

ion

seei

nter

actio

nlis

tlo

gfil

e11

9,17

3

M mak

end

x13

4,22

0m

ass,

mod

ified

seem

odifi

edm

ass

Max

wel

lian

dist

ribut

ion

17M

Dno

n-eq

uilib

rium∼

seen

on-e

quili

briu

mM

Dpa

ralle

l∼se

epar

alle

lMD

mdr

un22

1m

ean

forc

e,po

tent

ials

ofse

epot

entia

lsof

mea

nfo

rce

mec

hani

cs,s

tatis

tical

sees

tatis

tical

mec

hani

csm

emor

y,sh

ared

sees

hare

dm

emor

ym

esos

copi

cdy

nam

ics

2M

essa

geP

assi

ng37

Inte

rfac

ese

eMP

Im

irror

imag

e57

mk

angn

dx13

4,22

2m

odel

ing,

mol

ecul

arse

emol

ecul

arm

odel

ing

mod

ified

mas

s10

9m

olec

ular

mod

elin

g1

mot

ion,

equa

tions

ofse

eequ

atio

nsof

mot

ion

MP

I38

,42

N near

esti

mag

e18

neig

hbor

list

18,1

20N

eigh

bor

sear

chin

g18

,120

neig

hbor

,thi

rdse

ethi

rdne

ighb

orng

mx

135,

222

NM

Rre

finem

ent

127

62nm

run

34,2

23no

n-bo

nded

para

met

er84

Non

-equ

ilibr

ium

MD

128

15N

orm

alm

ode

anal

ysis

34N

ose-

Hoo

ver

tem

pera

ture

coup

ling

22nu

cleu

sse

epar

ticle

O octa

hedr

on13

onlin

em

anua

l11

5O

PLS

60op

tions

177,

223

P para

bolic

forc

e53

para

llelM

D42

Par

alle

l,D

ata

seeD

ata

Par

alle

lpa

ralle

lizat

ion

37pa

ram

eter

79bo

nded∼

seeb

onde

dpa

ram

eter

non-

bond

ed∼

seen

on-b

onde

dpa

ram

eter

Par

amet

er,R

unse

eRun

Par

amet

erP

arrin

ello

-Rah

man

pres

sure

coup

ling

24pa

rtic

le79

Par

ticle

Dec

ompo

sitio

n38

part

icle

-mes

hE

wal

dse

ePM

EP

artic

le-P

artic

leP

artic

le-M

esh

seeP

PP

Mpd

b2gm

x61

,82,

109,

224

perf

orm

ance

161

Per

iodi

cB

ound

ary

Con

ditio

ns15

7P

erio

dic

boun

dary

cond

ition

s11 74

plan

argr

oup

57P

ME

75,1

21P

oiss

onso

lver

53

v

Onl

ine

Res

ourc

es

You

can

find

mor

edo

cum

enta

tion

and

othe

rm

ater

iala

tour

hom

epag

ew

ww

.gro

mac

s.or

g.A

mon

got

her

thin

gsth

ere

isan

onlin

ere

fere

nce,

seve

ralG

RO

MA

CS

mai

ling

lists

with

arch

ives

and

con-

trib

uted

topo

logi

es/fo

rce

field

s.

Man

uala

ndG

RO

MA

CS

vers

ions

We

try

tore

leas

ean

upda

ted

vers

ion

ofth

em

anua

lwhe

neve

rw

ere

leas

ea

new

vers

ion

ofth

eso

ft-w

are,

soin

gene

rali

tis

ago

odid

eato

use

am

anua

lwith

the

sam

em

ajor

and

min

orre

leas

enu

mbe

ras

your

GR

OM

AC

Sin

stal

latio

n.A

nyre

visi

onnu

mbe

rs(li

ke3.

0.1)

are

how

ever

inde

pend

ent,

tom

ake

itpo

ssib

leto

impl

emen

tbug

fixes

and

man

uali

mpr

ovem

ents

ifne

cess

ary.

GR

OM

AC

Sis

Fre

eS

oftw

are

The

entir

eG

RO

MA

CS

pack

age

isav

aila

ble

unde

rth

eG

NU

Gen

eral

Pub

licLi

cens

e.T

his

mea

nsit’

sfr

eeas

infr

eesp

eech

,not

just

that

you

can

use

itw

ithou

tpay

ing

usm

oney

.F

orde

tails

,che

ckth

eC

OP

YIN

Gfil

ein

the

sour

ceco

deor

cons

ult

ww

w.g

nu.o

rg/c

opyl

eft/g

pl.h

tml.

The

GR

OM

AC

Sso

urce

code

and

Linu

xpa

ckag

esar

eav

aila

ble

onou

rho

mep

age!

Page 6: CS USER AL Simulations - GROMACS

viIn

dex

24

3

Freeze

group14

functionautocorrelation∼

seeautocorrelation

functionpotential∼

seepotentialfunction

shift∼se

eshiftfunctionTabulated∼

seeTabulated

function

Gganaeig

146,181g

analyze146,182

gangle

141,184g

bond141,185

gbundle

185g

chi186

gcluster

187g

com136

gconfrm

s189

gcoord

151g

covar146,189

gdensity

151,190g

dielectric191

gdih

191g

dipoles139,140,192

gdisre

193g

dist194

gdyndom

194g

enemat

195g

energy135,140,174,196

ggyrate

143,197g

h2order197

ghbond

147,198g

helix199

glie

200g

mdm

at143,201

gm

indist143,201

gm

orph202

gm

sd140,202

gnm

eig34,203

gnm

ens203

gorder

150,204g

potential151,204

gpvd

151g

rama

148,205g

rdf137,205

grm

s144,206

grm

sdist144,207

grm

sf208

grotacf

139,209g

runrm

s145

gsaltbr

210g

sas210

gsgangle

141,143,211g

sorient212

gtcaf

212g

traj213

gvelacc

139,214genbox

215genconf

216genion

216genpr

217gm

xcheck217

gmxdum

p218

GM

XR

C155

Grid

search19

gromos-87

47grom

os-96files

78force

field77

grompp

96,109,218groupaccelerate∼

seeaccelerate

groupcharge∼

seecharge

groupE

nergym

onitor∼se

eEnergy

monitor

groupF

reeze∼se

eFreeze

groupplanar∼

seeplanar

group

Hharmonic

interaction85

Hessian

34highw

ay220

htmlm

anual115

hydrogendatabase

88hydrogen-bond

81hypercube

38

Iimage,nearest

seenearestim

ageIm

properdihedral

5784index

file134

Page 7: CS USER AL Simulations - GROMACS

24

2In

dex

term

ini∼

seet

erm

inid

atab

ase

Dec

ompo

sitio

nP

artic

le∼

seeP

artic

leD

ecom

posi

tion

Spa

ce∼

seeS

pace

Dec

ompo

sitio

nde

gree

sof

free

dom

108

diel

ectr

icco

nsta

nt50

,121

diffu

sion

coef

ficie

nt14

0di

hedr

al57

Impr

oper∼

seeI

mpr

oper

dihe

dral

impr

oper∼

seei

mpr

oper

dihe

dral

Pro

per∼

seeP

rope

rdi

hedr

alpr

oper∼

seep

rope

rdi

hedr

aldi

sper

sion

48co

rrec

tion

122

Dis

tanc

ere

stra

int

62di

stan

cere

stra

ints

127

dist

ribut

ion,

Max

wel

lian

seeM

axw

ellia

ndi

strib

utio

ndo

dssp

148,

156,

178

dosh

ift15

1,15

6do

deca

hedr

on13

doub

lepr

ecis

ion

seep

reci

sion

,dou

ble

dum

my

seep

artic

leD

umm

yat

om71

81,1

08D

ynam

ics,

Bro

wni

anse

eBro

wni

anD

ynam

ics

dyna

mic

sLa

ngev

in∼

seeL

ange

vin

dyna

mic

sm

esos

copi

c∼se

emes

osco

pic

dyna

mic

sD

ynam

ics,

Sto

chas

ticse

eSto

chas

ticD

ynam

ics

dyna

mic

s,st

ocha

stic

sees

toch

astic

dyna

mic

s

E editc

onf

179

Ein

stei

nre

latio

n14

0E

lect

ricfie

ld12

9el

ectr

osta

ticfo

rce

19E

lect

rost

atic

s12

0en

econ

v18

0en

ergy

file

172

Ene

rgy

min

imiz

atio

n11

9

mon

itor

grou

p15

ener

gy kine

tic∼

seek

inet

icen

ergy

pote

ntia

l∼se

epot

entia

lene

rgy

ense

mbl

eav

erag

e1

equa

tion,

Sch

rodin

ger

seeS

chro

ding

ereq

uatio

neq

uatio

nsof

mot

ion

2,21

equi

libra

tion

173

esse

ntia

ldyn

amic

sse

ecov

aria

nce

anal

ysis

Ess

entia

lDyn

amic

sS

ampl

ing

36E

wal

dsu

m53

,74,

120

Ew

ald,

part

icle

-mes

h53

Exc

lusi

ons

85ex

clus

ions

69en

ergy

mon

itor

grou

p∼15

exte

nded

ense

mbl

e22

F File

type

115

file

ener

gy∼

seee

nerg

yfil

ein

dex∼

seei

ndex

file

log∼

seel

ogfil

eTo

polo

gy∼

seeT

opol

ogy

file

traj

ecto

ry∼

seet

raje

ctor

yfil

efil

es,g

rom

osse

egro

mos

-96

files

For

cefie

ld4

forc

e field

47,8

1C

onst

rain

t∼se

eCon

stra

intf

orce

cons

trai

nt∼

seec

onst

rain

tfor

ceel

ectr

osta

tic∼

seee

lect

rost

atic

forc

epa

rabo

lic∼

seep

arab

olic

forc

epo

tent

ialo

fmea

n∼se

epot

entia

lsof

mea

nfo

rce

forc

efiel

d,A

ll-hy

drog

ense

eAll-

hydr

ogen

forc

efiel

dF

ortr

an16

1F

ree

ener

gyca

lcul

atio

ns34

,99

Fre

een

ergy

free

ener

gyca

lcul

atio

ns10

3F

ree en

ergy

inte

ract

ions

66E

nerg

yP

ertu

rbat

ion

128

free

dom

,deg

rees

ofse

edeg

rees

offr

eedo

m

Con

tent

s

1In

trodu

ctio

n1

1.1

Com

puta

tiona

lChe

mis

try

and

Mol

ecul

arM

odel

ing..

..

..

..

..

..

..

..

1

1.2

Mol

ecul

arD

ynam

ics

Sim

ulat

ions.

..

..

..

..

..

..

..

..

..

..

..

..

.2

1.3

Ene

rgy

Min

imiz

atio

nan

dS

earc

hM

etho

ds..

..

..

..

..

..

..

..

..

..

.5

2D

efini

tions

and

Uni

ts7

2.1

Not

atio

n.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

7

2.2

MD

units

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

7

2.3

Red

uced

units

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

9

3A

lgor

ithm

s11

3.1

Intr

oduc

tion

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.11

3.2

Per

iodi

cbo

unda

ryco

nditi

ons..

..

..

..

..

..

..

..

..

..

..

..

..

..

11

3.2.

1S

ome

usef

ulbo

xty

pes.

..

..

..

..

..

..

..

..

..

..

..

..

..

13

3.2.

2C

ut-o

ffre

stric

tions

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.14

3.3

The

grou

pco

ncep

t..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.14

3.4

Mol

ecul

arD

ynam

ics

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

15

3.4.

1In

itial

cond

ition

s.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.15

3.4.

2N

eigh

bor

sear

chin

g..

..

..

..

..

..

..

..

..

..

..

..

..

..

.18

3.4.

3C

ompu

tefo

rces.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

20

3.4.

4U

pdat

eco

nfigu

ratio

n..

..

..

..

..

..

..

..

..

..

..

..

..

..

21

3.4.

5Te

mpe

ratu

reco

uplin

g..

..

..

..

..

..

..

..

..

..

..

..

..

.21

3.4.

6P

ress

ure

coup

ling.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.23

3.4.

7O

utpu

tste

p..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.26

3.5

She

llm

olec

ular

dyna

mic

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

28

3.5.

1O

ptim

izat

ion

ofth

esh

ellp

ositi

ons..

..

..

..

..

..

..

..

..

..

.28

Page 8: CS USER AL Simulations - GROMACS

viiiC

on

ten

ts

3.6C

onstraintalgorithms.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.28

3.6.1S

HA

KE

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.29

3.6.2LIN

CS

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.29

3.7S

imulated

Annealing.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.31

3.8S

tochasticD

ynamics.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.32

3.9B

rownian

Dynam

ics.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.32

3.10E

nergyM

inimization

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

33

3.10.1S

teepestDescent..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

33

3.10.2C

onjugateG

radient...

..

..

..

..

..

..

..

..

..

..

..

..

..

33

3.11N

ormalM

odeA

nalysis..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.34

3.12F

reeenergy

calculations..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

34

3.13E

ssentialDynam

icsS

ampling.

..

..

..

..

..

..

..

..

..

..

..

..

..

36

3.14P

arallelization..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

37

3.14.1M

ethodsofparallelization.

..

..

..

..

..

..

..

..

..

..

..

..

37

3.14.2M

Don

aring

ofprocessors...

..

..

..

..

..

..

..

..

..

..

..

39

3.15P

arallelMolecular

Dynam

ics..

..

..

..

..

..

..

..

..

..

..

..

..

..

42

3.15.1D

omain

decomposition.

..

..

..

..

..

..

..

..

..

..

..

..

..

42

3.15.2D

omain

decomposition

fornon-bonded

forces.

..

..

..

..

..

..

..

43

3.15.3P

arallelPP

PM

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.44

3.15.4P

arallelsorting..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.45

4F

orcefields

47

4.1N

on-bondedinteractions.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.48

4.1.1T

heLennard-Jones

interaction...

..

..

..

..

..

..

..

..

..

..

48

4.1.2B

uckinghampotential.

..

..

..

..

..

..

..

..

..

..

..

..

..

49

4.1.3C

oulomb

interaction..

..

..

..

..

..

..

..

..

..

..

..

..

..

50

4.1.4C

oulomb

interactionw

ithreaction

field..

..

..

..

..

..

..

..

..

50

4.1.5M

odifiednon-bonded

interactions..

..

..

..

..

..

..

..

..

..

.51

4.1.6M

odifiedshort-range

interactionsw

ithE

wald

summ

ation.

..

..

..

..

53

4.2B

ondedinteractions.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

54

4.2.1B

ondstretching.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

54

4.2.2M

orsepotentialbond

stretching..

..

..

..

..

..

..

..

..

..

..

55

4.2.3C

ubicbond

stretchingpotential.

..

..

..

..

..

..

..

..

..

..

.56

4.2.4H

armonic

anglepotential..

..

..

..

..

..

..

..

..

..

..

..

..

56

Index

τT

22εr

501-4

interaction59,85

Aaccelerategroup

15A

FM

pulling103

All-hydrogen

forcefield77

Am

dahl’slaw

38A

nglerestraint

62angle

vibration56

annealing,simulated

seesim

ulatedannealing

atomse

eparticletype

80D

umm

y∼se

eDum

my

atomdum

my∼

seedum

my

atomunited∼

seeunited

atomautocorrelation

function137

average,ensemble

seeensem

bleaverage

BBerendsen

temperature

coupling22

bondshellse

eparticlestretching

54bonded

parameter

83B

orn-Oppenheim

er4

Boundary

Conditions,P

eriodicse

ePeriodic

Boundary

Conditions

boundaryconditions,P

eriodicse

ePeriodic

boundaryconditions

Brow

nianD

ynamics

32B

uckingham49

buildingblock

82,86

Ccenter-of-mass

velocity17

Charge

Group

70

chargegroup

19,121chem

istry,computational

seecom

putationalchemistry

citingiv

coefficient,diffusionse

ediffusioncoefficient

combination

rule85

compressibility

24com

putationalchemistry

1C

onjugateG

radient33

conjugategradient

117connection

85constant,dielectric

seedielectric

constantC

onstraint28,85

constraint4

Constraintforce

100103

constraints126

convention,polymer

seepolym

erconvention

correlation137

Coulom

b50,66

couplingP

ressure∼se

ePressure

couplingS

urfacetension∼

seeS

urfacetension

couplingTem

perature∼se

eTemperature

couplingtem

perature∼se

etemperature

couplingC

ovarianceanalysis

145cut-off

51,70,121,122

DData

Parallel

37D

atabase86

databasehydrogen∼

seehydrogen

database

Page 9: CS USER AL Simulations - GROMACS

24

0B

iblio

gra

ph

yC

on

ten

tsix

4.2.

5C

osin

eba

sed

angl

epo

tent

ial

..

..

..

..

..

..

..

..

..

..

..

..

57

4.2.

6Im

prop

erdi

hedr

als.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

57

4.2.

7P

rope

rdi

hedr

als.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.57

4.2.

8S

peci

alin

tera

ctio

ns..

..

..

..

..

..

..

..

..

..

..

..

..

..

.60

4.2.

9P

ositi

onre

stra

ints.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.61

4.2.

10A

ngle

rest

rain

ts..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.62

4.2.

11D

ista

nce

rest

rain

ts..

..

..

..

..

..

..

..

..

..

..

..

..

..

.62

4.3

Fre

een

ergy

inte

ract

ions.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

66

4.3.

1S

oft-

core

inte

ract

ions.

..

..

..

..

..

..

..

..

..

..

..

..

..

.68

4.4

Met

hods

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.69

4.4.

1E

xclu

sion

san

d1-

4In

tera

ctio

ns...

..

..

..

..

..

..

..

..

..

..

69

4.4.

2C

harg

eG

roup

s...

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.70

4.4.

3T

reat

men

tofc

ut-o

ffs.

..

..

..

..

..

..

..

..

..

..

..

..

..

.70

4.5

Dum

my

atom

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

71

4.6

Long

Ran

geE

lect

rost

atic

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

.74

4.6.

1E

wal

dsu

mm

atio

n..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

74

4.6.

2P

ME

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

75

4.6.

3P

PP

M.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.75

4.6.

4O

ptim

izin

gF

ourie

rtr

ansf

orm

s..

..

..

..

..

..

..

..

..

..

..

.76

4.7

All-

hydr

ogen

forc

efiel

d..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.77

4.8

GR

OM

OS

-96

note

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.77

4.8.

1T

heG

RO

MO

S-9

6fo

rce

field.

..

..

..

..

..

..

..

..

..

..

..

.77

4.8.

2G

RO

MO

S-9

6fil

es.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

78

5To

polo

gies

79

5.1

Intr

oduc

tion

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.79

5.2

Par

ticle

type

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.79

5.2.

1A

tom

type

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.80

5.2.

2D

umm

yat

oms

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.81

5.3

Par

amet

erfil

es.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

82

5.3.

1A

tom

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

82

5.3.

2B

onde

dpa

ram

eter

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

.83

5.3.

3N

on-b

onde

dpa

ram

eter

s..

..

..

..

..

..

..

..

..

..

..

..

..

84

5.3.

41-

4in

tera

ctio

ns.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

85

Page 10: CS USER AL Simulations - GROMACS

xC

on

ten

ts

5.3.5E

xclusions.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

85

5.4C

onstraints.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

85

5.5D

atabases..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

86

5.5.1R

esiduedatabase.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.86

5.5.2H

ydrogendatabase.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

88

5.5.3Term

inidatabase..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

89

5.6F

ileform

ats.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

91

5.6.1Topology

file.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.91

5.6.2M

olecule.itpfile

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

96

5.6.3Ifdefoption

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.98

5.6.4F

reeenergy

calculations..

..

..

..

..

..

..

..

..

..

..

..

..

99

5.6.5C

onstraintforce..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.100

5.6.6C

oordinatefile

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.101

6S

pecialTopics103

6.1C

alculatingpotentials

ofmean

force:the

pullcode...

..

..

..

..

..

..

.103

6.1.1O

verview.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.103

6.1.2U

sage..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..104

6.1.3O

utput.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..107

6.1.4Lim

itations.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..107

6.1.5Im

plementation.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..108

6.1.6F

uturedevelopm

ent...

..

..

..

..

..

..

..

..

..

..

..

..

..108

6.2R

emoving

fastestdegreesoffreedom

..

..

..

..

..

..

..

..

..

..

..

..

108

6.2.1H

ydrogenbond-angle

vibrations...

..

..

..

..

..

..

..

..

..

.109

6.2.2O

ut-of-planevibrations

inarom

aticgroups..

..

..

..

..

..

..

..

110

6.3V

iscositycalculation

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..111

6.4U

serspecified

potentialfunctions...

..

..

..

..

..

..

..

..

..

..

..

.113

6.5R

unningG

RO

MA

CS

inparallel.

..

..

..

..

..

..

..

..

..

..

..

..

.114

7R

unparam

etersand

Program

s115

7.1O

nlineand

htmlm

anuals..

..

..

..

..

..

..

..

..

..

..

..

..

..

..115

7.2F

iletypes

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..115

7.3R

unP

arameters.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..117

7.3.1G

eneral..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.117

7.3.2P

reprocessing..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..117

Bib

liogra

ph

y2

39

[62]de

Loof,H

.,N

ilsson,L.,

Rigler,

R.

Molecular

dynamics

simulations

ofgalanin

inaqueous

andnonaqueous

solution.J.A

m.C

hem.S

oc.114:4028–4035,1992.

[63]F

eenstra,K

.A

.,H

ess,B

.,B

erendsen,H

.J.

C.

Improving

efficiencyof

largetim

e-scalem

oleculardynam

icssim

ulationsof

hydrogen-richsystem

s.J.

Com

p.C

hem.

20:786–798,1999.

[64]F

eenstra,K.A

.,Scheek,R

.M.,B

erendsen,H.J.C

.,Mark,A

.E.

Analysis

ofthehierarchy

ofm

otionofglobular

proteinsw

ithim

plicationsfor

proteinfolding.

(submitted

toP

RO

TE

INS

:S

truct.Funct.G

en.,feb.2001).

[65]A

llen,M

.P.,

Tildesley,

D.

J.C

omputer

Sim

ulationsof

Liquids.O

xford:O

xfordS

cienceP

ublications.1987.

[66]van

derS

poel,D

.,B

erendsen,H

.J.

C.

Molecular

dynamics

simulations

ofLeu-enkephalin

inw

aterand

DM

SO

.B

iophys.J.72:2032–2041,1997.

[67]van

derS

poel,D.,van

Maaren,P.J.,B

erendsen,H.J.C

.A

systematic

studyofw

aterm

odelsfor

molecular

simulation.

J.Chem

.Phys.108:10220–10230,1998.

[68]S

mith,P.E

.,vanG

unsteren,W.F.

The

viscosityofspc

andspc/e

water.

Com

p.Phys.C

omm

.215:315–318,1993.

[69]B

alasubramanian,S

.,Mundy,C

.J.,Klein,M

.L.S

hearviscosity

ofpolarfluids:

Miolecular

dynamics

calculationsofw

ater.J.C

hem.P

hys.105:11190–11195,1996.

[70]A

madei,

A.,

Linssen,A

.B

.M

.,B

erendsen,H

.J.

C.

Essentialdynam

icsof

proteins.P

RO

-T

EIN

S:S

truct.Funct.G

en.17:412–425,1993.

[71]H

ess,B.

Sim

ilaritiesbetw

eenprincipalcom

ponentsofprotein

dynamics

andrandom

diffu-sion.

Phys.R

ev.E62:8438–8448,2000.

[72]K

absch,W

.,S

ander,C

.D

ictionaryof

proteinsecondary

structure:P

atternrecognition

ofhydrogen-bonded

andgeom

etricalfeatures.B

iopolymers

22:2577–2637,1983.

[73]W

illiamson,M

.P.,Asakura,T.

Em

piricalcomparisons

ofmodels

forchem

ical-shiftcalcula-tion

inproteins.

J.Magn.R

eson.Ser.B

101:63–71,1993.

[74]B

ekker,H

.,B

erendsen,H

.J.

C.,

Dijkstra,

E.

J.,A

chterop,S

.,v.

Drunen,

R.,

v.d.

Spoel,

D.,

Sijbers,

A.,

Keegstra,

H.,

Reitsm

a,B

.,R

enardus,M

.K

.R

.G

romacs

method

ofvirial

calculationusing

asingle

sum.

InPhysics

Co

mp

utin

g9

2(Singapore,1993).

deG

root,R.A

.,N

adrchal,J.,eds..W

orldS

cientific.

[75]B

erendsen,H

.J.

C.,

Grigera,

J.R

.,S

traatsma,

T.P.

The

missing

termin

effectivepair

potentials.J.P

hys.Chem

.91:6269–6271,1987.

[76]B

ekker,H

.O

ntwerp

vaneen

special-purposecom

putervoor

moleculaire

dynamica

simu-

laties.M

aster’sthesis.

RuG

.1987.

[77]van

Gunsteren,W

.F.,Berendsen,H

.J.C.

Moleculardynam

icsofsim

plesystem

s.P

racticumH

andleidingvoorM

DP

racticumN

ijenborgh4,9747

AG

,Groningen,T

heN

etherlands1994.

Page 11: CS USER AL Simulations - GROMACS

23

8B

iblio

gra

ph

y

[46]

van

Gun

ster

en,

W.

F.,

Bill

eter

,S

.R

.,E

isin

g,A

.A

.,H

unen

berg

er,

P.H

.,K

ruger

,P.

,M

ark,

A.E

.,S

cott,

W.R

.P.,

Tiro

ni,I

.G.B

iom

olec

ular

Sim

ulat

ion:

The

GR

OM

OS

96m

anua

land

user

guid

e.Zu

rich,

Sw

itzer

land

:H

ochs

chul

verla

gA

Gan

der

ET

HZ

uric

h.19

96.

[47]

Mor

se,

P.M

.D

iato

mic

mol

ecul

esac

cord

ing

toth

ew

ave

mec

hani

cs.

II.vi

brat

iona

llev

els.

Phy

s.R

ev.3

4:57

–64,

1929

.

[48]

Ber

ends

en,H

.J.C

.,P

ostm

a,J.

P.M

.,va

nG

unst

eren

,W.F

.,H

erm

ans,

J.In

tera

ctio

nm

odel

sfo

rw

ater

inre

latio

nto

prot

ein

hydr

atio

n.In

:In

term

olec

ular

For

ces.

Pul

lman

,B

.ed

..

D.

Rei

delP

ublis

hing

Com

pany

Dor

drec

ht19

8133

1–34

2.

[49]

Fer

guso

n,D

.M.

Par

amet

rizat

ion

and

eval

uatio

nof

afle

xibl

ew

ater

mod

el.

J.C

omp.

Che

m.

16:5

01–5

11,1

995.

[50]

Jorg

ense

n,W

.L.

,T

irado

-Riv

es,

J.T

heO

PLS

pote

ntia

lfun

ctio

nsfo

rpr

otei

ns.

ener

gym

in-

imiz

atio

nsfo

rcr

ysta

lsof

cycl

icpe

ptid

esan

dcr

ambi

n.J.

Am

.C

hem

.S

oc.

110:

1657

–166

6,19

88.

[51]

Tord

a,A

.E

.,S

chee

k,R

.M

.,va

nG

unst

eren

,W

.F.

Tim

e-de

pend

ent

dist

ance

rest

rain

tsin

mol

ecul

ardy

nam

ics

sim

ulat

ions

.C

hem

.Phy

s.Le

tt.15

7:28

9–29

4,19

89.

[52]

van

Gun

ster

en,W

.F.,

Mar

k,A

.E.

Valid

atio

nof

mol

ecul

ardy

nam

ics

sim

ulat

ions

.J.

Che

m.

Phy

s.10

8:61

09–6

116,

1998

.

[53]

Ber

ends

en,H

.J.C

.,va

nG

unst

eren

,W.F

.M

olec

ular

dyna

mic

ssi

mul

atio

ns:

Tech

niqu

esan

dap

proa

ches

.In

:M

olec

ular

Liqu

ids-

Dyn

amic

san

dIn

tera

ctio

ns.e

tal.,

A.J

.B.e

d.N

ATO

AS

IC

135.

Rei

delD

ordr

echt

,The

Net

herla

nds

1984

475–

500.

[54]

Ew

ald,

P.P.

Die

Ber

echn

ung

optis

cher

und

elek

tros

tatis

cher

Gitt

erpo

tent

iale

.A

nn.

Phy

s.64

:253

–287

,192

1.

[55]

Dar

den,

T.,

York

,D

.,P

eder

sen,

L.P

artic

lem

esh

Ew

ald:

An

N-lo

g(N

)m

etho

dfo

rE

wal

dsu

ms

inla

rge

syst

ems.

J.C

hem

.Phy

s.98

:100

89–1

0092

,199

3.

[56]

Ess

man

n,U

.,P

erer

a,L.

,B

erko

witz

,M

.L.

,D

arde

n,T.

,Le

e,H

.,P

eder

sen,

L.G

.A

smoo

thpa

rtic

lem

esh

ewal

dpo

tent

ial.

J.C

hem

.Phy

s.10

3:85

77–8

592,

1995

.

[57]

Hoc

kney

,R.W

.,E

astw

ood,

J.W

.Com

pute

rsim

ulat

ion

usin

gpa

rtic

les.

New

York

:M

cGra

w-

Hill

.198

1.

[58]

Luty

,B.A

.,T

ironi

,I.G

.,va

nG

unst

eren

,W.F

.La

ttice

-sum

met

hods

for

calc

ulat

ing

elec

tro-

stat

icin

tera

ctio

nsin

mol

ecul

arsi

mul

atio

ns.

J.C

hem

.Phy

s.10

3:30

14–3

021,

1995

.

[59]

Kin

g,P.

M.,

Mar

k,A

.E.,

van

Gun

ster

en,W

.F.

Re-

para

met

eriz

atio

nof

arom

atic

inte

ract

ions

inth

eG

RO

MO

Sfo

rce-

field

.P

rivat

eC

omm

unic

atio

n19

93.

[60]

Ryc

kaer

t,J.

P.,B

elle

man

s,A

.F

ar.D

isc.

Che

m.S

oc.6

6:95

,197

8.

[61]

onB

ioch

emic

alN

omen

clat

ure,

I.-I.

C.

Abr

revi

atio

nsan

dsy

mbo

lsfo

rth

ede

scrip

tion

ofth

eco

nfor

mat

ion

ofpo

lype

ptid

ech

ains

.ten

tativ

eru

les

(196

9).

Bio

chem

istr

y9:

3471

–347

8,19

70.

Co

nte

nts

xi

7.3.

3R

unco

ntro

l..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.117

7.3.

4La

ngev

indy

nam

ics.

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

18

7.3.

5E

nerg

ym

inim

izat

ion.

..

..

..

..

..

..

..

..

..

..

..

..

..

.119

7.3.

6S

hell

Mol

ecul

arD

ynam

ics.

..

..

..

..

..

..

..

..

..

..

..

..

119

7.3.

7O

utpu

tcon

trol

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.119

7.3.

8N

eigh

bor

sear

chin

g..

..

..

..

..

..

..

..

..

..

..

..

..

..

.120

7.3.

9E

lect

rost

atic

san

dV

dW.

..

..

..

..

..

..

..

..

..

..

..

..

..1

20

7.3.

10Te

mpe

ratu

reco

uplin

g..

..

..

..

..

..

..

..

..

..

..

..

..

.123

7.3.

11P

ress

ure

coup

ling.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.123

7.3.

12S

imul

ated

anne

alin

g..

..

..

..

..

..

..

..

..

..

..

..

..

..1

25

7.3.

13Ve

loci

tyge

nera

tion.

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

25

7.3.

14B

onds

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

26

7.3.

15E

nerg

ygr

oup

excl

usio

ns..

..

..

..

..

..

..

..

..

..

..

..

..1

27

7.3.

16N

MR

refin

emen

t..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

27

7.3.

17F

ree

Ene

rgy

Per

turb

atio

n..

..

..

..

..

..

..

..

..

..

..

..

..1

28

7.3.

18N

on-e

quili

briu

mM

D.

..

..

..

..

..

..

..

..

..

..

..

..

..

.128

7.3.

19E

lect

ricfie

lds.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.129

7.3.

20U

ser

defin

edth

ingi

es..

..

..

..

..

..

..

..

..

..

..

..

..

..1

29

7.4

Pro

gram

sby

topi

c..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

29

8A

naly

sis

133

8.1

Gro

ups

inA

naly

sis..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

33

8.1.

1D

efau

ltG

roup

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.134

8.2

Look

ing

atyo

urtr

ajec

tory

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.135

8.3

Gen

eral

prop

ertie

s..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

35

8.4

Rad

iald

istr

ibut

ion

func

tions

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

36

8.5

Cor

rela

tion

func

tions

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

37

8.5.

1T

heor

yof

corr

elat

ion

func

tions

..

..

..

..

..

..

..

..

..

..

..

137

8.5.

2U

sing

FF

Tfo

rco

mpu

tatio

nof

the

AC

F..

..

..

..

..

..

..

..

..

139

8.5.

3S

peci

alfo

rms

ofth

eA

CF.

..

..

..

..

..

..

..

..

..

..

..

..

.139

8.5.

4S

ome

App

licat

ions

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.139

8.5.

5M

ean

Squ

are

Dis

plac

emen

t..

..

..

..

..

..

..

..

..

..

..

..

140

8.6

Bon

ds,a

ngle

san

ddi

hedr

als.

..

..

..

..

..

..

..

..

..

..

..

..

..

.140

8.7

Rad

ius

ofgy

ratio

nan

ddi

stan

ces..

..

..

..

..

..

..

..

..

..

..

..

..1

43

Page 12: CS USER AL Simulations - GROMACS

xiiC

on

ten

ts

8.8R

ootmean

squaredeviations

instructure

..

..

..

..

..

..

..

..

..

..

..

144

8.9C

ovarianceanalysis.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..145

8.10H

ydrogenbonds.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..147

8.11P

roteinrelated

items.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.148

8.12Interface

relateditem

s...

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.150

8.13C

hemicalshifts.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..151

ATechnicalD

etails153

A.1

Installation.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..153

A.2

Single

orD

oubleprecision.

..

..

..

..

..

..

..

..

..

..

..

..

..

..153

A.3

Porting

GR

OM

AC

S.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..154

A.3.1

Multi-processor

Optim

ization..

..

..

..

..

..

..

..

..

..

..

.155

A.4

Environm

entVariables..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.155

BS

ome

implem

entationdetails

157

B.1

Single

Sum

Virialin

GR

OM

AC

S..

..

..

..

..

..

..

..

..

..

..

..

..

157

B.1.1

Virial.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..157

B.1.2

Virialfrom

non-bondedforces...

..

..

..

..

..

..

..

..

..

..

.158

B.1.3

The

intramolecular

shift(mol-shift)..

..

..

..

..

..

..

..

..

..

.158

B.1.4

Virialfrom

CovalentB

onds...

..

..

..

..

..

..

..

..

..

..

..

159

B.1.5

Virialfrom

Shake.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.160

B.2

Optim

izations.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.160

B.2.1

InnerLoops

forW

ater..

..

..

..

..

..

..

..

..

..

..

..

..

.160

B.2.2

Fortran

Code.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.161

B.3

Com

putationofthe

1.0/sqrtfunction..

..

..

..

..

..

..

..

..

..

..

..

.161

B.3.1

Introduction..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.161

B.3.2

General.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..161

B.3.3

Applied

tofloating

pointnumbers.

..

..

..

..

..

..

..

..

..

..

162

B.3.4

Specification

ofthelookup

table..

..

..

..

..

..

..

..

..

..

..

163

B.3.5

Separate

exponentandfraction

computation

..

..

..

..

..

..

..

..

164

B.3.6

Implem

entation..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.165

B.4

Tabulatedfunctions.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..165

CLong

rangecorrections

167

C.1

Dispersion.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.167

Bib

liogra

ph

y2

37

[30]Levitt,M

.,Sander,C

.,Stern,P.S

.T

henorm

almodes

ofaprotein:

Native

bovinepancreatic

trypsininhibitor.

Proc.N

atl.Acad.S

ci.US

A10:181–199,1983.

[31]G

o,N

.,N

oguti,T.,

Nishikaw

a,T.

Dynam

icsof

asm

allglobular

proteinin

terms

oflow

-frequency

vibrationalmodes.

Proc.N

atl.Acad.S

ci.US

A80:3696–3700,1983.

[32]B

rooks,B.,K

arplus,M.

Harm

onicdynam

icsofproteins:

Norm

almodes

andfluctuations

inbovine

pancreatictrypsin

inhibitor.P

roc.Natl.A

cad.Sci.U

SA

80:6571–6575,1983.

[33]H

ayward,S

.,Go,N.

Collective

variabledescription

ofnativeprotein

dynamics.

Annu.R

ev.P

hys.Chem

.46:223–250,1995.

[34]de

Groot,B

.L.,Am

adei,A.,van

Aalten,D

.M.F.,B

erendsen,H.J.C

.Towards

anexhaustive

sampling

oftheconfigurationalspaces

ofthetw

oform

softhe

peptidehorm

oneguanylin.

J.B

iomol.S

tr.Dyn.13(5):741–751,1996.

[35]de

Groot,

B.

L.,A

madei,

A.,

Scheek,

R.

M.,

vanN

uland,N

.A

.J.,

Berendsen,

H.

J.C

.A

nextended

sampling

ofthe

configurationalspaceof

hprfrome.

coli.

PR

OT

EIN

S:

Struct.

Funct.G

en.26:314–322,1996.

[36]V

riend,G

.W

HAT

IF:

am

olecularm

odelingand

drugdesign

program.

J.M

ol.G

raph.8:52–56,1990.

[37]F

incham,D

.P

arallelcomputers

andm

olecularsim

ulation.M

ol.Sim

.1:1,1987.

[38]R

aine,A

.R

.C

.,F

incham,

D.,

Sm

ith,W

.S

ystolicloop

methods

form

oleculardynam

icssim

ulation.C

omp.P

hys.Com

m.55:13–30,1989.

[39]van

Gunsteren,

W.

F.,B

erendsen,H

.J.

C.

Grom

os-87m

anual.B

iomos

BV

Nijenborgh

4,9747

AG

Groningen,T

heN

etherlands1987.

[40]van

Buuren,

A.

R.,

Marrink,

S.

J.,B

erendsen,H

.J.

C.

Am

oleculardynam

icsstudy

ofthe

decane/water

interface.J.P

hys.Chem

.97:9206–9212,1993.

[41]M

ark,A

.E

.,van

Helden,

S.

P.,S

mith,

P.E

.,Janssen,

L.H

.M

.,van

Gunsteren,

W.

F.C

on-vergence

propertiesoffree

energycalculations:

α-cyclodextrin

complexes

asa

casestudy.

J.A

m.C

hem.S

oc.116:6293–6302,1994.

[42]Jorgensen,W

.L.,Chandrasekhar,J.,M

adura,J.D.,Im

pey,R.W

.,Klein,M

.L.C

omparison

ofsimple

potentialfunctionsfor

simulating

liquidw

ater.J.C

hem.P

hys.79:926–935,1983.

[43]van

Buuren,

A.

R.,

Berendsen,

H.

J.C

.M

oleculardynam

icssim

ulationof

thestability

ofa

22residue

alpha-helixin

water

and30

%trifluoroethanol.

Biopolym

ers33:1159–1166,

1993.

[44]Liu,

H.,

Muller-P

lathe,F.,

vanG

unsteren,W

.F.

Aforce

fieldfor

liquiddim

ethylsulfox-

ideand

liquidproporties

ofliquid

dimethylsulfoxide

calculatedusing

molecular

dynamics

simulation.

J.Am

.Chem

.Soc.117:4363–4366,1995.

[45]T

ironi,I.

G.,

Sperb,

R.,

Sm

ith,P.

E.,

vanG

unsteren,W

.F.

Ageneralized

reactionfield

method

form

oleculardynam

icssim

ulations.J.C

hem.P

hys.102:5451–5459,1995.

Page 13: CS USER AL Simulations - GROMACS

23

6B

iblio

gra

ph

y

[13]

Ber

ends

en,

H.

J.C

.E

lect

rost

atic

inte

ract

ions

.In

:C

ompu

ter

Sim

ulat

ion

ofB

iom

olec

ular

Sys

tem

s.va

nG

unst

eren

,W

.F.

,W

eine

r,P.

K.,

Wilk

inso

n,A

.J.

eds.

.E

SC

OM

Leid

en19

9316

1–18

1.

[14]

Hoc

kney

,R.W

.,G

oel,

S.P

.J.

Com

p.P

hys.

14:1

48,1

974.

[15]

Verle

t.,L.

Phy

s.R

ev.3

4:13

11–1

327,

1967

.

[16]

Ber

ends

en,H

.J.C

.,va

nG

unst

eren

,W.F

.P

ract

ical

algo

rithm

sfo

rdy

nam

ics

sim

ulat

ions

.

[17]

Ber

ends

en,

H.

J.C

.,P

ostm

a,J.

P.M

.,D

iNol

a,A

.,H

aak,

J.R

.M

olec

ular

dyna

mic

sw

ithco

uplin

gto

anex

tern

alba

th.

J.C

hem

.Phy

s.81

:368

4–36

90,1

984.

[18]

Nos

e,S

.A

mol

ecul

ardy

nam

ics

met

hod

for

sim

ulat

ions

inth

eca

noni

cale

nsem

ble.

Mol

.P

hys.

52:2

55–2

68,1

984.

[19]

Hoo

ver,

W.

G.

Can

onic

aldy

nam

ics:

equi

libriu

mph

ase-

spac

edi

strib

utio

ns.

Phy

s.R

ev.

A31

:169

5–16

97,1

985.

[20]

Ber

ends

en,H

.J.C

.Tra

nspo

rtpr

oper

ties

com

pute

dby

linea

rres

pons

eth

roug

hw

eak

coup

ling

toa

bath

.In

:C

ompu

ter

Sim

ulat

ions

inM

ater

ial

Sci

ence

.M

eyer

,M

.,P

ontik

is,

V.ed

s..

Klu

wer

1991

139–

155.

[21]

Par

rinel

lo,

M.,

Rah

man

,A

.P

olym

orph

ictr

ansi

tions

insi

ngle

crys

tals

:A

new

mol

ecul

ardy

nam

ics

met

hod.

J.A

ppl.

Phy

s.52

:718

2–71

90,1

981.

[22]

Nos

e,S

.,K

lein

,M

.L.

Con

stan

tpr

essu

rem

olec

ular

dyna

mic

sfo

rm

olec

ular

syst

ems.

Mol

.P

hys.

50:1

055–

1076

,198

3.

[23]

Dic

k,B

.G

.,O

verh

ause

r,A

.W

.T

heor

yof

the

diel

ectr

icco

nsta

nts

ofal

kali

halid

ecr

ysta

ls.

Phy

s.R

ev.1

12:9

0–10

3,19

58.

[24]

Jord

an,

P.C

.,va

nM

aare

n,P.

J.,

Mav

ri,J.

,va

nde

rS

poel

,D

.,B

eren

dsen

,H

.J.

C.

Tow

ards

phas

etr

ansf

erab

lepo

tent

ialf

unct

ions

:M

etho

dolo

gyan

dap

plic

atio

nto

nitr

ogen

.J.

Che

m.

Phy

s.10

3:22

72–2

285,

1995

.

[25]

van

Maa

ren,

P.J.

,van

der

Spo

el,D

.M

olec

ular

dyna

mic

ssi

mul

atio

nsof

aw

ater

with

ano

vel

shel

l-mod

elpo

tent

ial.

J.P

hys.

Che

m.B

.105

:261

8–26

26,2

001.

[26]

Ryc

kaer

t,J.

P.,

Cic

cotti

,G

.,B

eren

dsen

,H

.J.

C.

Num

eric

alin

tegr

atio

nof

the

cart

esia

neq

uatio

nsof

mot

ion

ofa

syst

emw

ithco

nstr

aint

s;m

olec

ular

dyna

mic

sof

n-al

kane

s.J.

Com

p.P

hys.

23:3

27–3

41,1

977.

[27]

Miy

amot

o,S

.,K

ollm

an,P

.A.

SE

TT

LE:A

nan

alyt

ical

vers

ion

ofth

eS

HA

KE

and

RAT

TLE

algo

rithm

sfo

rrig

idw

ater

mod

els.

J.C

omp.

Che

m.1

3:95

2–96

2,19

92.

[28]

Hes

s,B

.,B

ekke

r,H

.,B

eren

dsen

,H

.J.

C.,

Fra

aije

,J.

G.

E.

M.

LIN

CS

:A

linea

rco

nstr

aint

solv

erfo

rm

olec

ular

sim

ulat

ions

.J.

Com

p.C

hem

.18:

1463

–147

2,19

97.

[29]

van

Gun

ster

en,

W.

F.,

Ber

ends

en,

H.

J.C

.A

leap

-fro

gal

gorit

hmfo

rst

ocha

stic

dyna

mic

s.M

ol.S

im.1

:173

–185

,198

8.

Co

nte

nts

xiii

C.1

.1E

nerg

y.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

67

C.1

.2V

irial

and

pres

sure

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.168

DA

vera

ges

and

fluct

uatio

ns17

1

D.1

For

mul

aefo

rav

erag

ing.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

71

D.2

Impl

emen

tatio

n.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

72

D.2

.1P

arto

faS

imul

atio

n.

..

..

..

..

..

..

..

..

..

..

..

..

..

.173

D.2

.2C

ombi

ning

two

sim

ulat

ions

..

..

..

..

..

..

..

..

..

..

..

..

173

D.2

.3S

umm

ing

ener

gyte

rms.

..

..

..

..

..

..

..

..

..

..

..

..

..1

74

EM

anua

lPag

es17

7

E.1

optio

ns.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.177

E.2

dods

sp.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

78

E.3

editc

onf

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.179

E.4

enec

onv

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.180

E.5

gan

aeig

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.181

E.6

gan

alyz

e.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.182

E.7

gan

gle

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.184

E.8

gbo

nd.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.185

E.9

gbu

ndle

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.185

E.1

0g

chi.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

86

E.1

1g

clus

ter.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

87

E.1

2g

conf

rms

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

89

E.1

3g

cova

r.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

89

E.1

4g

dens

ity.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.190

E.1

5g

diel

ectr

ic.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

91

E.1

6g

dih

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.191

E.1

7g

dipo

les

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

92

E.1

8g

disr

e.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.193

E.1

9g

dist

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

94

E.2

0g

dynd

om.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.194

E.2

1g

enem

at.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.195

E.2

2g

ener

gy.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

96

E.2

3g

gyra

te.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

97

E.2

4g

h2or

der

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..1

97

Page 14: CS USER AL Simulations - GROMACS

xivC

on

ten

ts

E.25

ghbond

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.198

E.26

ghelix

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..199

E.27

glie

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.200

E.28

gm

dmat.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..201

E.29

gm

indist.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.201

E.30

gm

orph.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..202

E.31

gm

sd.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.202

E.32

gnm

eig.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..203

E.33

gnm

ens.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..203

E.34

gorder

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.204

E.35

gpotential.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.204

E.36

gram

a.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.205

E.37

grdf

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.205

E.38

grm

s.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.206

E.39

grm

sdist.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.207

E.40

grm

sf.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.208

E.41

grotacf

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.209

E.42

gsaltbr

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.210

E.43

gsas.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..210

E.44

gsgangle

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..211

E.45

gsorient.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..212

E.46

gtcaf

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..212

E.47

gtraj.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..213

E.48

gvelacc

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.214

E.49

genbox..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..215

E.50

genconf..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.216

E.51

genion..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..216

E.52

genpr.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.217

E.53

gmxcheck.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.217

E.54

gmxdum

p.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.218

E.55

grompp

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.218

E.56

highway

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.220

E.57

makendx

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..220

E.58

mdrun

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..221

Bibliography

[1]B

erendsen,H

.J.

C.,

vander

Spoel,

D.,

vanD

runen,R

.G

RO

MA

CS

:A

message-passing

parallelmolecular

dynamics

implem

entation.C

omp.P

hys.Com

m.91:43–56,1995.

[2]Lindahl,

E.,

Hess,

B.,

vander

Spoel,

D.

Grom

acs3.0:

Apackage

form

olecularsim

ulationand

trajectoryanalysis.

Inpress

2001.

[3]van

derS

poel,D

.,Vogel,

H.

J.,B

erendsen,H

.J.

C.

Molecular

dynamics

simulations

ofN

-terminal

peptidesfrom

anucleotide

bindingprotein.

PR

OT

EIN

S:

Struct.

Funct.

Gen.

24:450–466,1996.

[4]van

Gunsteren,

W.

F.,B

erendsen,H

.J.

C.

Com

putersim

ulationof

molecular

dynamics:

Methodology,

applications,and

perspectivesin

chemistry.

Angew

.C

hem.

Int.E

d.E

ngl.29:992–1023,1990.

[5]F

raaije,J.G.E

.M.

Dynam

icdensity

functionaltheoryfor

microphase

separationkinetics

ofblock

copolymer

melts.

J.Chem

.Phys.99:9202–9212,1993.

[6]M

cQuarrie,D

.A.S

tatisticalMechanics.

New

York:H

arper&

Row

.1976.

[7]van

Gunsteren,

W.

F.,B

erendsen,H

.J.

C.

Algorithm

sfor

macrom

oleculardynam

icsand

constraintdynamics.

Mol.P

hys.34:1311–1327,1977.

[8]N

ilges,M.,C

lore,G.M

.,Gronenborn,A

.M.

Determ

inationofthree-dim

ensionalstructuresofproteins

frominterproton

distancedata

bydynam

icalsimulated

annealingfrom

arandom

arrayofatom

s.F

EB

SLett.239:129–136,1988.

[9]van

Schaik,

R.

C.,

Berendsen,

H.

J.C

.,Torda,

A.

E.,

vanG

unsteren,W

.F.

Astructure

refinement

method

basedon

molecular

dynamics

in4

spatialdim

ensions.J.

Mol.

Biol.

234:751–762,1993.

[10]Z

imm

erman,K

.Allpurpose

molecularm

echanicssim

ulatorandenergy

minim

izer.J.Com

p.C

hem.12:310–319,1991.

[11]A

dams,

D.

J.,A

dams,

E.

M.,

Hills,

G.

J.T

hecom

putersim

ulationof

polarliquids.

Mol.

Phys.38:387–400,1979.

[12]B

ekker,H

.,D

ijkstra,E

.J.,

Renardus,

M.

K.

R.,

Berendsen,

H.

J.C

.A

nefficient,

boxshape

independentnon-bonded

forceand

virialalgorithm

form

oleculardynam

ics.M

ol.S

im.14:137–152,1995.

Page 15: CS USER AL Simulations - GROMACS

23

4A

pp

en

dix

E.

Ma

nu

alP

age

sC

on

ten

tsxv

E.5

9m

kan

gndx

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.222

E.6

0ng

mx

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

22

E.6

1nm

run

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

23

E.6

2op

tions

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

23

E.6

3pd

b2gm

x.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.224

E.6

4pr

oton

ate

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

25

E.6

5tp

bcon

v.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

26

E.6

6tr

jcat

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.226

E.6

7tr

jcon

v.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.227

E.6

8tr

jord

er.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

28

E.6

9w

heel

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

29

E.7

0x2

top

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

29

E.7

1xm

drun

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.230

E.7

2xp

m2p

s.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

31

E.7

3xr

ama

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..2

32

Bib

liogr

aphy

235

Inde

x24

1

Page 16: CS USER AL Simulations - GROMACS

xviC

on

ten

tsE

.73

.xra

ma

23

3

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

Page 17: CS USER AL Simulations - GROMACS

23

2A

pp

en

dix

E.

Ma

nu

alP

age

s

sele

cted

toco

mbi

neth

em

atric

es.

Inth

isca

se,

ane

wco

lor

map

will

bege

nera

ted

with

are

dgr

adie

ntfo

rne

gativ

enu

mbe

rsan

da

blue

for

posi

tive.

Ifth

eco

lor

codi

ngan

dle

gend

labe

lsof

both

mat

rices

are

iden

tical

,onl

yon

ele

gend

will

bedi

spla

yed,

else

two

sepa

rate

lege

nds

are

disp

laye

d.

-titl

eca

nbe

set

tono

ne

tosu

ppre

ssth

etit

le,

ortoy

lab

el

tosh

owth

etit

lein

the

Y-la

belp

ositi

on(a

long

side

the

Y-ax

is).

With

the

-ra

inb

ow

optio

ndu

llgr

ey-s

cale

mat

rices

can

betu

rned

into

attr

activ

eco

lor

pict

ures

.

Mer

ged

orra

inbo

wed

mat

rices

can

bew

ritte

nto

anX

Pix

elM

apfil

ew

ithth

e-x

pm

optio

n.

File

s-f

roo

t.xp

mIn

put

XP

ixM

apco

mpa

tible

mat

rixfil

e-f

2ro

ot2

.xp

mIn

put,

Opt

.X

Pix

Map

com

patib

lem

atrix

file

-di

ps.

m2

pIn

put,

Opt

.,Li

b.Inpu

tfile

for

mat

2ps

-do

ou

t.m

2p

Out

put,

Opt

.In

putfi

lefo

rm

at2p

s-o

plo

t.e

ps

Out

put,

Opt

.E

ncap

sula

ted

Pos

tScr

ipt(

tm)

file

-xp

mro

ot.xp

mO

utpu

t,O

pt.

XP

ixM

apco

mpa

tible

mat

rixfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-fra

me

bool

yes

Dis

play

fram

e,tic

ks,l

abel

s,tit

lean

dle

gend

-titl

een

umto

pS

how

title

at:to

p,o

nce

,yla

be

lor

no

ne

-yo

nce

bool

no

Sho

wy-

labe

lonl

yon

ce-le

ge

nd

enum

bo

thS

how

lege

nd:b

oth

,first

,se

con

dor

no

ne

-dia

gen

umfir

stD

iago

nal:f

irst

,se

con

dor

no

ne

-co

mb

ine

enum

ha

lve

sC

ombi

netw

om

atric

es:ha

lve

s,a

dd

,su

b,m

ult

ord

iv-b

xre

al0

Box

x-si

ze(a

lso

y-si

zew

hen

-by

isno

tset

)-b

yre

al0

Box

y-si

ze-r

ain

bo

wen

umn

oR

ainb

owco

lors

,con

vert

whi

teto

:no,b

lue

orre

d-g

rad

ien

tve

ctor

00

0R

e-sc

ale

colo

rmap

toa

smoo

thgr

adie

ntfr

omw

hite

1,1,

1to

r,g,

b-s

kip

int

1on

lyw

rite

oute

very

nr-t

hro

wan

dco

lum

n-z

ero

line

bool

no

inse

rtlin

ein

xpm

mat

rixw

here

axis

labe

lis

zero

E.7

3xr

ama

xram

ash

ows

aR

amac

hand

ran

mov

ie,

that

is,

itsh

ows

the

Phi

/Psi

angl

esas

afu

nctio

nof

time

inan

X-

Win

dow

.

Sta

ticP

hi/P

sipl

ots

for

prin

ting

can

bem

ade

with

gra

ma.

Som

eof

the

mor

eco

mm

onX

com

man

dlin

eop

tions

can

beus

ed:

-bg,

-fg

chan

geco

lors

,-fo

ntfo

ntna

me,

chan

ges

the

font

.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l

List

ofF

igur

es

3.1

Per

iodi

cbo

unda

ryco

nditi

ons

intw

odi

men

sion

s...

..

..

..

..

..

..

..

.12

3.2

Arh

ombi

cdo

deca

hedr

onan

dtr

unca

ted

octa

hedr

on(a

rbitr

ary

orie

ntat

ions

)..

..

13

3.3

The

glob

alM

Dal

gorit

hm.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.16

3.4

AM

axw

ellia

ndi

strib

utio

n,ge

nera

ted

from

rand

omnu

mbe

rs.

..

..

..

..

..

.17

3.5

Grid

sear

chin

two

dim

ensi

ons.

The

arro

ws

are

the

box

vect

ors.

..

..

..

..

..

19

3.6

The

Leap

-Fro

gin

tegr

atio

nm

etho

d...

..

..

..

..

..

..

..

..

..

..

..

.21

3.7

The

MD

upda

teal

gorit

hm.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

27

3.8

The

thre

epo

sitio

nup

date

sne

eded

for

one

time

step

..

..

..

..

..

..

..

..

.30

3.9

Fre

een

ergy

cycl

es..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

35

3.10

The

inte

ract

ion

mat

rix..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

40

3.11

Inte

ract

ion

mat

rices

for

diffe

rentN

..

..

..

..

..

..

..

..

..

..

..

..

.40

3.12

The

Par

alle

lMD

algo

rithm

...

..

..

..

..

..

..

..

..

..

..

..

..

..

.41

3.13

Dat

aflo

win

arin

gof

proc

esso

rs.

..

..

..

..

..

..

..

..

..

..

..

..

..

42

3.14

Inde

xin

the

coor

dina

tear

ray..

..

..

..

..

..

..

..

..

..

..

..

..

..

.43

4.1

The

Lenn

ard-

Jone

sin

tera

ctio

n...

..

..

..

..

..

..

..

..

..

..

..

..

.48

4.2

The

Buc

king

ham

inte

ract

ion..

..

..

..

..

..

..

..

..

..

..

..

..

..

.49

4.3

The

Cou

lom

bin

tera

ctio

nw

ithan

dw

ithou

trea

ctio

nfie

ld.

..

..

..

..

..

..

.50

4.4

The

Cou

lom

bF

orce

,Shi

fted

For

cean

dS

hift

Fun

ctio

nS

(r),

..

..

..

..

..

..

53

4.5

Bon

dst

retc

hing

...

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.54

4.6

The

Mor

sepo

tent

ialw

ell,

with

bond

leng

th0.

15nm

...

..

..

..

..

..

..

.56

4.7

Ang

levi

brat

ion.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.57

4.8

Impr

oper

dihe

dral

angl

es...

..

..

..

..

..

..

..

..

..

..

..

..

..

..

58

4.9

Impr

oper

dihe

dral

pote

ntia

l...

..

..

..

..

..

..

..

..

..

..

..

..

..

58

4.10

Pro

per

dihe

dral

angl

e...

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

59

4.11

Ryc

kaer

t-B

elle

man

sdi

hedr

alpo

tent

ial.

..

..

..

..

..

..

..

..

..

..

..

.60

Page 18: CS USER AL Simulations - GROMACS

xviiiL

istofF

igu

res

4.12P

ositionrestraintpotential..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

61

4.13D

istanceR

estraintpotential...

..

..

..

..

..

..

..

..

..

..

..

..

..

63

4.14S

oft-coreinteractions

atλ

=0.5,w

ithC

A6=C

A12

=C

B6=C

B12

=1.

..

..

..

.68

4.15A

toms

alongan

alkanechain.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

69

4.16D

umm

yatom

construction....

..

..

..

..

..

..

..

..

..

..

..

..

..

72

6.1S

chematic

pictureofpulling

alipid

outofalipid

bilayerw

ithA

FM

pulling.V

rup

isthe

velocityatw

hichthe

springis

retracted,Z

link

isthe

atomto

which

thespring

isattached

andZsp

rin

gis

thelocation

ofthespring..

..

..

..

..

..

..

..

.104

6.2O

verviewof

thedifferent

referencegroup

possibilities,applied

tointerface

sys-tem

s.C

isthe

referencegroup.

The

circlesrepresent

thecenter

ofm

assof

2groups

plusthe

referencegroup,and

dc

isthe

referencedistance...

..

..

..

.105

6.3D

umm

yatom

constructionsfor

hydrogenatom

s..

..

..

..

..

..

..

..

..

.109

6.4D

umm

yatom

constructionsfor

aromatic

residues..

..

..

..

..

..

..

..

..

111

8.1T

hew

indowofn

gm

xshow

inga

boxofw

ater...

..

..

..

..

..

..

..

..

.136

8.2D

efinitionofslices

ingrd

f..

..

..

..

..

..

..

..

..

..

..

..

..

..

.137

8.3gO

O(r)

forO

xygen-Oxygen

ofSP

C-w

ater....

..

..

..

..

..

..

..

..

..

138

8.4M

eanS

quareD

isplacementofS

PC

-water.

..

..

..

..

..

..

..

..

..

..

.141

8.5D

ihedralconventions...

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..142

8.6O

ptionsofg

sga

ng

le.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..142

8.7A

minim

umdistance

matrix

fora

peptide[

3]..

..

..

..

..

..

..

..

..

..

144

8.8G

eometricalH

ydrogenbond

criterion....

..

..

..

..

..

..

..

..

..

..

147

8.9Insertion

ofwater

intoan

H-bond..

..

..

..

..

..

..

..

..

..

..

..

..

147

8.10A

nalysisofthe

secondarystructure

elements

ofapeptide

intim

e..

..

..

..

..

149

8.11D

efinitionofthe

dihedralanglesφandψ

oftheprotein

backbone...

..

..

..

149

8.12R

amachandran

plotofasm

allprotein..

..

..

..

..

..

..

..

..

..

..

..

149

8.13H

elicalwheelprojection

oftheN

-terminalhelix

ofHP

r..

..

..

..

..

..

..

.150

B.1

IEE

Esingle

precisionfloating

pointformat.

..

..

..

..

..

..

..

..

..

.162

E.7

2.

xpm

2p

s2

31

theusualoutputw

illbew

rittento

file.W

henrunning

with

MP

I,asignalto

oneofthe

mdrun

processesis

sufficient,thissignalshould

notbesentto

mpirun

orthe

mdrun

processthatis

theparentofthe

others.

Files

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-otra

j.trrO

utputF

ullprecisiontrajectory:

trrtrj

-xtra

j.xtcO

utput,Opt.

Com

pressedtrajectory

(portablexdr

format)

-cco

nfo

ut.g

roO

utputG

enericstructure:

grog96

pdb-e

en

er.e

dr

Output

Generic

energy:edr

ene-g

md

.log

Output

Logfile

-dg

dl

dg

dl.xvg

Output,O

pt.xvgr/xm

grfile

-tab

leta

ble

.xvgInput,O

pt.xvgr/xm

grfile

-reru

nre

run

.xtcInput,O

pt.G

enerictrajectory:

xtctrr

trjgrog96

pdb-e

isa

m.e

di

Input,Opt.

ED

sampling

input-e

osa

m.e

do

Output,O

pt.E

Dsam

plingoutput

-jw

ha

m.g

ctInput,O

pt.G

eneralcouplingstuff

-job

am

.gct

Input,Opt.

Generalcoupling

stuff-ffo

ut

gct.xvg

Output,O

pt.xvgr/xm

grfile

-de

vou

td

evia

tie.xvg

Output,O

pt.xvgr/xm

grfile

-run

av

run

ave

r.xvgO

utput,Opt.

xvgr/xmgr

file-p

ip

ull.p

pa

Input,Opt.

Pullparam

eters-p

op

ullo

ut.p

pa

Output,O

pt.P

ullparameters

-pd

pu

ll.pd

oO

utput,Opt.

Pulldata

output-p

np

ull.n

dx

Input,Opt.

Indexfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-de

ffnm

stringS

etthedefaultfilenam

efor

allfileoptions

-np

int1

Num

berofnodes,m

ustbethe

same

asused

forgrom

pp-v

booln

oB

eloud

andnoisy

-com

pa

ctbool

yes

Write

acom

pactlogfile

-mu

ltibool

no

Do

multiple

simulations

inparallel(only

with

-np>1)

-gla

sbool

no

Do

glasssim

ulationw

ithspeciallong

rangecorrections

-ion

izebool

no

Do

asim

ulationincluding

theeffectofan

X-R

aybom

bardmenton

yoursystem

E.72

xpm2ps

xpm2ps

makes

abeautifulcolor

plotofanX

PixelM

apfile.

Labelsand

axiscan

bedisplayed,w

henthey

aresupplied

inthe

correctm

atrixform

at.M

atrixdata

may

begenerated

byprogram

ssuch

asdo

dssp,grm

sor

gm

dmat.

Param

etersare

setinthem2

pfile

optionallysupplied

with-d

i.

Reasonable

defaultsare

provided.S

ettingsforthe

y-axisdefaultto

thoseforthe

x-axis.F

ontnames

havea

defaultinghierarchy:

titlefont->

legendfont;titlefont->

(xfont->yfont->

ytickfont)->

xtickfont,e.g.setting

titlefontsetsallfonts,setting

xfontsetsyfont,ytickfontand

xtickfont.

With

-f2a

2ndm

atrixfile

canbe

supplied,both

matrix

filesw

illberead

simultaneously

andthe

upperleft

halfof

thefirst

one(-f

)is

plottedtogether

with

thelow

erright

halfof

thesecond

one(

-f2).

The

diagonalwillcontain

valuesfrom

them

atrixfile

selectedw

ith-d

iag

.P

lottingofthe

diagonalvaluescan

besuppressed

altogetherby

setting-d

iag

ton

on

e.

With

-com

bin

ean

alternativeoperation

canbe

Page 19: CS USER AL Simulations - GROMACS

23

0A

pp

en

dix

E.

Ma

nu

alP

age

s

-kt

real

40

0A

ngle

forc

eco

nsta

nt(k

J/m

ol/r

ad2 )-k

pre

al5

Dih

edra

lang

lefo

rce

cons

tant

(kJ/

mol

/rad2 )

-ne

xcl

int

3N

umbe

rof

excl

usio

ns-H

14

bool

no

Use

3rd

neig

hbou

rin

tera

ctio

nsfo

rhy

drog

enat

oms

-alld

ihbo

oln

oG

ener

ate

allp

rope

rdi

hedr

als

-ro

un

dbo

olye

sR

ound

offm

easu

red

valu

es-p

airs

bool

yes

Out

put1

-4in

tera

ctio

ns(p

airs

)in

topo

logy

file

-na

me

strin

gIC

EN

ame

ofyo

urm

olec

ule

•T

heat

omty

pese

lect

ion

ispr

imiti

ve.

Virt

ually

noch

emic

alkn

owle

dge

isus

ed

•P

erio

dic

boun

dary

cond

ition

ssc

rew

upth

ebo

ndin

g

•N

oim

prop

erdi

hedr

als

are

gene

rate

d

•T

heat

oms

toat

omty

petr

ansl

atio

nta

ble

isin

com

plet

e(f

fG43

a1.n

2tfil

ein

the

$GM

XLI

Bdi

rect

ory)

.P

leas

eex

tend

itan

dse

ndth

ere

sults

back

toth

eG

RO

MA

CS

crew

.

E.7

1xm

drun

xmdr

unis

the

expe

rimen

talM

Dpr

ogra

m.

New

feat

ures

are

test

edin

this

prog

ram

befo

rebe

ing

impl

emen

ted

inth

ede

faul

tm

drun

.C

urre

ntly

unde

rin

vest

igat

ion

are:

pola

rizib

ility

,gl

ass

sim

ulat

ions

,F

ree

ener

gype

r-tu

rbat

ion,

X-R

aybo

mba

rdm

ents

and

para

lleli

ndep

ende

ntsi

mul

atio

ns.It

read

sth

eru

nin

put

file

(-s

)an

ddi

strib

utes

the

topo

logy

over

node

sif

need

ed.

The

coor

dina

tes

are

pass

edar

ound

,so

that

com

puta

tions

can

begi

n.F

irsta

neig

hbor

listi

sm

ade,

then

the

forc

esar

eco

mpu

ted.

The

forc

esar

egl

obal

lysu

mm

ed,a

ndth

eve

loci

ties

and

posi

tions

are

upda

ted.

Ifne

cess

ary

shak

eis

perf

orm

edto

cons

trai

nbo

ndle

ngth

san

d/or

bond

angl

es.

Tem

pera

ture

and

Pre

ssur

eca

nbe

cont

rolle

dus

ing

wea

kco

uplin

gto

aba

th.

mdr

unpr

oduc

esat

leas

tthr

eeou

tput

file,

plus

one

log

file

(-g

)pe

rno

de.

The

traj

ecto

ryfil

e(

-o),

cont

ains

coor

dina

tes,

velo

citie

san

dop

tiona

llyfo

rces

.T

hest

ruct

ure

file

(-c

)co

ntai

nsth

eco

ordi

nate

san

dve

loci

ties

ofth

ela

stst

ep.

The

ener

gyfil

e(

-e)

cont

ains

ener

gies

,the

tem

pera

ture

,pre

ssur

e,et

c,a

loto

fthe

seth

ings

are

also

prin

ted

inth

elo

gfil

eof

node

0.O

ptio

nally

coor

dina

tes

can

bew

ritte

nto

aco

mpr

esse

dtr

ajec

tory

file

(-x

).

Whe

nru

nnin

gin

para

llelw

ithP

VM

oran

old

vers

ion

ofM

PIt

he-np

optio

nm

ustb

egi

ven

toin

dica

teth

enu

mbe

rof

node

s.

The

optio

n-d

gd

lis

only

used

whe

nfr

eeen

ergy

pert

urba

tion

istu

rned

on.

With

-re

run

anin

putt

raje

ctor

yca

nbe

give

nfo

rw

hich

forc

esan

den

ergi

esw

illbe

(re)

calc

ulat

ed.

Nei

gh-

bor

sear

chin

gw

illbe

perf

orm

edfo

rev

ery

fram

e,un

less

nst

list

isze

ro(s

eeth

e.md

pfil

e).

ED

(ess

entia

ldyn

amic

s)sa

mpl

ing

issw

itche

don

byus

ing

the

-ei

flag

follo

wed

byan

.ed

ifil

e.T

he.e

di

file

can

bepr

oduc

edus

ing

optio

nsin

the

essd

ynm

enu

ofth

eW

HAT

IFpr

ogra

m.

mdr

unpr

oduc

esa

.ed

ofil

eth

atco

ntai

nspr

ojec

tions

ofpo

sitio

ns,v

eloc

ities

and

forc

eson

tose

lect

edei

genv

ecto

rs.

The

-tab

leop

tion

can

beus

edto

pass

mdr

una

form

atte

dta

ble

with

user

-defi

ned

pote

ntia

lfun

ctio

ns.

The

file

isre

adfr

omei

ther

the

curr

entd

irect

ory

orfr

omth

eG

MX

LIB

dire

ctor

y.A

num

ber

ofpr

efor

mat

ted

tabl

esar

epr

esen

ted

inth

eG

MX

LIB

dir,

for

6-8,

6-9,

6-10

,6-

11,

6-12

Lenn

ard

Jone

spo

tent

ials

with

norm

alC

oulo

mb.

The

optio

ns-p

i,-

po

,-p

d,-

pn

are

used

for

pote

ntia

lofm

ean

forc

eca

lcul

atio

nsan

dum

brel

lasa

mpl

ing.

See

man

ual.

Whe

nm

drun

rece

ives

aT

ER

Msi

gnal

,itw

illse

tnst

eps

toth

ecu

rren

tste

ppl

uson

e.W

hen

mdr

unre

ceiv

esa

US

R1

sign

al,

itw

illse

tns

teps

toth

ene

xtm

ultip

leof

nstx

out

afte

rth

ecu

rren

tst

ep.

Inbo

thca

ses

all

List

ofTa

bles

1.1

Typi

calv

ibra

tiona

lfre

quen

cies

...

..

..

..

..

..

..

..

..

..

..

..

..

.3

2.1

Bas

icun

itsus

edin

GR

OM

AC

S..

..

..

..

..

..

..

..

..

..

..

..

..

.8

2.2

Der

ived

units

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

8

2.3

Som

eP

hysi

calC

onst

ants.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.9

2.4

Red

uced

Lenn

ard-

Jone

squ

antit

ies

..

..

..

..

..

..

..

..

..

..

..

..

.9

3.1

The

cubi

cbo

x,th

erh

ombi

cdo

deca

hedr

onan

dth

etr

unca

ted

octa

hedr

on.

..

..

.13

3.2

The

num

ber

ofin

tera

ctio

nsbe

twee

npa

rtic

les...

..

..

..

..

..

..

..

..

.40

4.1

Con

stan

tsfo

rR

ycka

ert-

Bel

lem

ans

pote

ntia

l(kJ

mol

−1).

..

..

..

..

..

..

..

59

4.2

Par

amet

ers

for

the

diffe

rent

func

tiona

lfor

ms

ofth

eno

n-bo

nded

inte

ract

ions

..

..

71

5.1

Par

ticle

type

sin

GR

OM

AC

S..

..

..

..

..

..

..

..

..

..

..

..

..

..

80

5.2

Sta

ticat

omty

pepr

oper

ties

inG

RO

MA

CS.

..

..

..

..

..

..

..

..

..

..

83

5.3

The

topo

logy

(*.to

p)

file.

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.92

5.4

The

mol

ecul

ede

finiti

on..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

.93

7.1

The

GR

OM

AC

Sfil

ety

pes

and

how

they

can

beus

edfo

rin

put/o

utpu

t..

..

..

.11

6

Page 20: CS USER AL Simulations - GROMACS

xxL

istofTa

ble

sE

.69

.w

he

el

22

9

ofthefirstn

waters

ism

ade,theordered

trajectorycan

beused

with

anyG

romacs

programto

analyzethe

nclosestw

aters.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-n

ind

ex.n

dx

Input,Opt.

Indexfile

-oo

rde

red

.xtcO

utputG

enerictrajectory:

xtctrr

trjgrog96

pdb

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-na

int3

Num

berofatom

sin

am

olecule-d

aint

1A

tomused

forthe

distancecalculation

E.69

wheel

wheelplots

ahelicalw

heelrepresentationofyour

sequence.The

inputsequenceis

inthe

.datfilew

herethe

firstlinecontains

thenum

berofresidues

andeach

consecutiveline

containsa

residuename.

Files

-fn

nn

ice.d

at

InputG

enericdata

file-o

plo

t.ep

sO

utputE

ncapsulatedP

ostScript(tm

)file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-r0int

1T

hefirstresidue

number

inthe

sequence-ro

t0real

0R

otatearound

anangle

initially(90

degreesm

akessense)

-Tstring

Plota

titlein

thecenter

ofthew

heel(mustbe

shorterthan

10characters,

oritw

illoverwrite

thew

heel)-n

nbool

yes

Togglenum

bers

E.70

x2top

x2topgenerates

aprim

itivetopology

froma

coordinatefile.

The

programassum

esallhydrogens

arepresent

when

definingthe

hybridizationfrom

theatom

name

andthe

number

ofbonds.T

heprogram

canalso

make

anrtp

entry,which

youcan

thenadd

tothe

rtpdatabase.

Files

-fco

nf.g

roInput

Generic

structure:gro

g96pdb

tprtpb

tpa-o

ou

t.top

Output,O

pt.Topology

file-r

ou

t.rtpO

utput,Opt.

Residue

Typefile

usedby

pdb2gmx

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int0

Setthe

nicelevel-kb

real4

00

00

0B

ondedforce

constant(kJ/mol/nm 2)

Page 21: CS USER AL Simulations - GROMACS

22

8A

pp

en

dix

E.

Ma

nu

alP

age

s

Usi

ng-t

run

ctr

jcon

vca

ntr

unca

te.tr

jin

plac

e,i.e

.w

ithou

tcop

ying

the

file.

Thi

sis

usef

ulw

hen

aru

nha

scr

ashe

ddu

ring

disk

I/O(o

nem

ore

disk

full)

,or

whe

ntw

oco

ntig

uous

traj

ecto

ries

mus

tbe

conc

aten

ated

with

outh

ave

doub

lefr

ames

.

trjc

at

ism

ore

suita

ble

for

conc

aten

atin

gtr

ajec

tory

files

.

Opt

ion

-du

mp

can

beus

edto

extr

acta

fram

eat

orne

aron

esp

ecifi

ctim

efr

omyo

urtr

ajec

tory

.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-otr

ajo

ut.xt

cO

utpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put,

Opt

.S

truc

ture

+m

ass(

db):

tpr

tpb

tpa

gro

g96

pdb

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e-f

rfr

am

es.

nd

xIn

put,

Opt

.In

dex

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-t

uen

ump

sT

ime

unit:

ps

,fs

,ns

,us

,ms,

s,m

orh

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-s

kip

int

1O

nly

writ

eev

ery

nr-t

hfr

ame

-dt

time

0O

nly

writ

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-du

mp

time

-1D

ump

fram

ene

ares

tspe

cifie

dtim

e(p

s)-t

0tim

e0

Sta

rtin

gtim

e(p

s)(d

efau

lt:do

n’tc

hang

e)-t

ime

ste

ptim

e0

Cha

nge

time

step

betw

een

inpu

tfra

mes

(ps)

-pb

cen

umn

on

eP

BC

trea

tmen

t:no

ne

,wh

ole

,in

bo

xor

no

jum

p-u

ren

umre

ctU

nit-

cell

repr

esen

tatio

n:re

ct,t

ric

orco

mp

act

-ce

nte

rbo

oln

oC

ente

rat

oms

inbo

x-b

ox

vect

or0

00

Siz

efo

rne

wcu

bic

box

(def

ault:

read

from

inpu

t)-s

hift

vect

or0

00

All

coor

dina

tes

will

besh

ifted

byfr

amen

r*sh

ift-f

itbo

oln

oF

itm

olec

ule

tore

fstr

uctu

rein

the

stru

ctur

efil

e-p

fitbo

oln

oP

rogr

essi

vefit

,to

the

prev

ious

fitte

dst

ruct

ure

-nd

ec

int

3P

reci

sion

for

.xtc

and

.gro

writ

ing

innu

mbe

rof

deci

mal

plac

es-v

el

bool

yes

Rea

dan

dw

rite

velo

citie

sif

poss

ible

-fo

rce

bool

no

Rea

dan

dw

rite

forc

esif

poss

ible

-tru

nc

time

-1T

runc

ate

inpu

ttrj

file

afte

rth

istim

e(p

s)-e

xec

strin

gE

xecu

teco

mm

and

for

ever

you

tput

fram

ew

ithth

efr

ame

num

ber

asar

-gu

men

t-a

pp

bool

no

App

end

outp

ut-s

plit

time

0S

tart

writ

ing

new

file

whe

ntM

OD

split

=fir

sttim

e(p

s)-s

ep

bool

no

Writ

eea

chfr

ame

toa

sepa

rate

.gro

or.p

dbfil

e

E.6

8tr

jord

er

trjo

rder

orde

rsm

olec

ules

acco

rdin

gto

the

smal

lest

dist

ance

toat

oms

ina

refe

renc

egr

oup.

Itw

illas

kfo

ra

grou

pof

refe

renc

eat

oms

and

agr

oup

ofm

olec

ules

.F

orea

chfr

ame

ofth

etr

ajec

tory

the

sele

cted

mol

ecul

esw

illbe

reor

dere

dac

cord

ing

toth

esh

orte

stdi

stan

cebe

twee

nat

omnu

mbe

r-d

ain

the

mol

ecul

ean

dal

lthe

atom

sin

the

refe

renc

egr

oup.

All

atom

sin

the

traj

ecto

ryar

ew

ritte

nto

the

outp

uttr

ajec

tory

.

trjo

rder

can

beus

eful

for

e.g.

anal

yzin

gth

en

wat

ers

clos

est

toa

prot

ein.

Inth

atca

seth

ere

fere

nce

grou

pw

ould

beth

epr

otei

nan

dth

egr

oup

ofm

olec

ules

wou

ldco

nsis

tofa

llth

ew

ater

atom

s.W

hen

anin

dex

grou

p

Cha

pter

1

Intro

duct

ion

1.1

Com

puta

tiona

lChe

mis

try

and

Mol

ecul

arM

odel

ing

GR

OM

AC

Sis

anen

gine

tope

rfor

mm

olec

ular

dyna

mic

ssi

mul

atio

nsan

den

ergy

min

imiz

atio

n.T

hese

are

two

ofth

em

any

tech

niqu

esth

atbe

long

toth

ere

alm

ofco

mpu

tatio

nalc

hem

istr

yan

dm

olec

ular

mod

elin

g.Co

mp

uta

tion

alC

he

mis

tryis

just

ana

me

toin

dica

teth

eus

eof

com

puta

tiona

lte

chni

ques

inch

emis

try,

rang

ing

from

quan

tum

mec

hani

csof

mol

ecul

esto

dyna

mic

sof

larg

eco

mpl

exm

olec

ular

aggr

egat

es.

Mo

lecu

lar

mo

de

lingi

ndic

ates

the

gene

ralp

roce

ssof

desc

ribin

gco

mpl

exch

emic

alsy

stem

sin

term

sof

are

alis

ticat

omic

mod

el,

with

the

aim

toun

ders

tand

and

pred

ictm

acro

scop

icpr

oper

ties

base

don

deta

iled

know

ledg

eon

anat

omic

scal

e.O

ften

mol

ecul

arm

odel

ing

isus

edto

desi

gnne

wm

ater

ials

,for

whi

chth

eac

cura

tepr

edic

tion

ofph

ysic

alpr

oper

ties

ofre

alis

ticsy

stem

sis

requ

ired.

Mac

rosc

opic

phys

ical

prop

ertie

sca

nbe

dist

ingu

ishe

din

(a)

sta

tice

qu

ilib

riu

mp

rop

ert

ies,s

uch

asth

ebi

ndin

gco

nsta

ntof

anin

hibi

tor

toan

enzy

me,

the

aver

age

pote

ntia

len

ergy

ofa

syst

em,

orth

era

dial

dist

ribut

ion

func

tion

ina

liqui

d,an

d(

b)d

yna

mic

or

no

n-e

qu

ilib

riu

mp

rop

ert

ies,su

chas

the

visc

osity

ofa

liqui

d,di

ffusi

onpr

oces

ses

inm

embr

anes

,th

edy

nam

ics

ofph

ase

chan

ges,

reac

tion

kine

tics,

orth

edy

nam

ics

ofde

fect

sin

crys

tals

.T

hech

oice

ofte

chni

que

depe

nds

onth

equ

estio

nas

ked

and

onth

efe

asib

ility

ofth

em

etho

dto

yiel

dre

liabl

ere

sults

atth

epr

esen

tsta

teof

the

art.

Idea

lly,t

he(r

elat

ivis

tic)

time-

depe

nden

tSch

rod

inge

req

uatio

nde

scrib

esth

epr

oper

ties

ofm

olec

ular

syst

ems

with

high

accu

racy

,bu

tan

ythi

ngm

ore

com

plex

than

the

equi

libriu

mst

ate

ofa

few

atom

sca

nnot

beha

ndle

dat

this

ab

initi

ole

vel.

Thu

sap

prox

imat

ions

are

man

dato

ry;

the

high

erth

eco

mpl

exity

ofa

syst

eman

dth

elo

nger

the

time

span

ofth

epr

oces

ses

ofin

tere

stis

,the

mor

ese

vere

appr

oxim

atio

nsar

ere

quire

d.A

tace

rtai

npo

int(

reac

hed

very

muc

hea

rlier

than

one

wou

ldw

ish)

thea

bin

itio

appr

oach

mus

tbe

augm

ente

dor

repl

aced

bye

mp

iric

alp

aram

eter

izat

ion

ofth

em

odel

used

.W

here

sim

ulat

ions

base

don

phys

ical

prin

cipl

esof

atom

icin

tera

ctio

nsst

illfa

ildu

eto

the

com

plex

ityof

the

syst

em(a

sis

unfo

rtun

atel

yst

illth

eca

sefo

rth

epr

edic

tion

ofpr

otei

nfo

ldin

g;bu

t:th

ere

isho

pe!)

mol

ecul

arm

odel

ing

isba

sed

entir

ely

ona

sim

ilarit

yan

alys

isof

know

nst

ruct

ural

and

chem

ical

data

.T

heQ

SA

Rm

etho

ds(Q

uant

itativ

eS

truc

ture

-Act

ivity

Rel

atio

ns)

and

man

yho

mol

ogy-

base

dpr

otei

nst

ruct

ure

pred

ictio

nsbe

long

toth

ela

tter

cate

gory

.

Mac

rosc

opic

prop

ertie

sar

eal

way

sen

sem

ble

aver

ages

over

are

pres

enta

tive

stat

istic

alen

sem

ble

Page 22: CS USER AL Simulations - GROMACS

2C

ha

pte

r1

.In

trod

uctio

n

(eitherequilibrium

ornon-equilibrium

)of

molecular

systems.

For

molecular

modeling

thishas

two

importantconsequences:

•T

heknow

ledgeof

asingle

structure,even

ifit

isthe

structureof

theglobal

energym

in-im

um,

isnot

sufficient.It

isnecessary

togenerate

arepresentative

ensemble

ata

giventem

perature,inorder

tocom

putem

acroscopicproperties.

Butthis

isnotenough

tocom

putetherm

odynamic

equilibriumproperties

thatare

basedon

freeenergies,

suchas

phaseequi-

libria,bindingconstants,solubilities,relative

stabilityofm

olecularconform

ations,etc.T

hecom

putationof

freeenergies

andtherm

odynamic

potentialsrequires

specialextensionsof

molecular

simulation

techniques.

•W

hilem

olecularsim

ulationsin

principleprovide

atomic

detailsof

thestructures

andm

o-tions,

suchdetails

areoften

notrelevant

forthe

macroscopic

propertiesof

interest.T

hisopens

thew

ayto

simplify

thedescription

ofinteractionsand

averageover

irrelevantdetails.T

hescience

ofstatistical

mechanics

providesthe

theoreticalfram

ework

forsuch

simpli-

fications.T

hereis

ahierarchy

ofm

ethodsranging

fromconsidering

groupsof

atoms

asone

unit,describing

motion

ina

reducednum

berof

collectivecoordinates,

averagingover

solventm

oleculesw

ithpotentials

ofm

eanforce

combined

with

stochasticdynam

ics[

4],to

me

sosco

pic

dyn

am

icsdescribingdensities

ratherthan

atoms

andfluxes

asresponse

totherm

odynamic

gradientsrather

thanvelocities

oraccelerations

asresponse

toforces

[5].

For

thegeneration

ofarepresentative

equilibriumensem

bletw

om

ethodsare

available:(

a)M

on

teC

arlo

simu

latio

nsand

(b)Mo

lecu

larD

yna

mics

simu

latio

ns.F

orthegeneration

ofnon-equilibriumensem

blesand

forthe

analysisofdynam

icevents,

onlythe

secondm

ethodis

appropriate.W

hileM

onteC

arlosim

ulationsare

more

simple

thanM

D(they

donotrequire

thecom

putationofforces),

theydo

notyieldsignificantly

betterstatisticsthan

MD

ina

givenam

ountofcomputertim

e.T

here-fore

MD

isthe

more

universaltechnique.Ifa

startingconfiguration

isvery

farfrom

equilibrium,

theforces

may

beexcessively

largeand

theM

Dsim

ulationm

ayfail.

Inthose

casesa

robuste

ne

rgy

min

imiza

tionis

required.A

notherreason

toperform

anenergy

minim

izationis

therem

ovalofallkinetic

energyfrom

thesystem

:if

several’snapshots’fromdynam

icsim

ulationsm

ustbe

com-

pared,energym

inimization

reducesthe

thermal’noise’in

thestructures

andpotentialenergies,so

thattheycan

becom

paredbetter.

1.2M

olecularD

ynamics

Sim

ulations

MD

simulations

solveN

ewton’s

equationsofm

otionfor

asystem

ofN

interactingatom

s:

mi ∂

2ri

∂t 2

=F

i ,i=

1...N

.(1.1)

The

forcesare

thenegative

derivativesofa

potentialfunctionV

(r1 ,r

2 ,...,rN

):

Fi =

−∂V

∂r

i(1.2)

The

equationsare

solvedsim

ultaneouslyin

small

time

steps.T

hesystem

isfollow

edfor

some

time,

takingcare

thatthe

temperature

andpressure

remain

atthe

requiredvalues,

andthe

coor-dinates

arew

rittento

anoutput

fileat

regularintervals.

The

coordinatesas

afunction

oftim

e

E.6

7.

trjconv

22

7

E.67

trjconv

trjconvcan

converttrajectoryfiles

inm

anyw

ays:1.from

oneform

attoanother

2.selectasubsetofatom

s3.rem

oveperiodicity

fromm

olecules4.keep

multim

ericm

oleculestogether

5.centeratom

sin

thebox

6.fitatoms

toreference

structure7.reduce

thenum

beroffram

es8.change

thetim

estamps

ofthefram

es(

-t0and

-time

step

)

The

programtrjca

tcan

concatenatem

ultipletrajectory

files.

Currently

sevenform

atsare

supportedfor

inputandoutput:

.xtc,.trr

,.trj,.g

ro,.g

96

,.pd

band

.g8

7.

The

fileform

atsare

detectedfrom

thefile

extension.T

heprecision

of.xtc

and.g

rooutput

istaken

fromthe

inputfilefor.xtc

,.g

roand

.pd

b,

andfrom

the-nd

ec

optionfor

otherinputform

ats.T

heprecision

isalw

aystaken

from-nd

ec

,when

thisoption

isset.

Allother

formats

havefixed

precision..trr

and.trj

outputcanbe

singleor

doubleprecision,depending

onthe

precisionofthe

trjconvbinary.

Note

thatvelocitiesare

onlysupported

in.trr

,.trj,.g

roand

.g9

6files.

Option

-ap

pcan

beused

toappend

outputtoan

existingtrajectory

file.N

ochecks

areperform

edto

ensureintegrity

ofthe

resultingcom

binedtrajectory

file..pdb

filesw

ithallfram

esconcatenated

canbe

viewed

with

rasm

ol

-nm

rpd

b.

Itis

possibleto

selectpart

ofyour

trajectoryand

write

itout

toa

newtrajectory

filein

orderto

savedisk

space,e.g.

forleaving

outthe

water

froma

trajectoryof

aprotein

inw

ater.A

LWAY

Sput

theoriginal

trajectoryon

tape!W

erecom

mend

touse

theportable

.xtcform

atfor

youranalysis

tosave

diskspace

andto

haveportable

files.

There

aretw

ooptions

forfitting

thetrajectory

toa

referenceeither

foressentialdynam

icsanalysis

orfor

whatever.

The

firstoptionis

justplainfitting

toa

referencestructure

inthe

structurefile,the

secondoption

isa

progressivefit

inw

hichthe

firsttim

eframe

isfitted

tothe

referencestructure

inthe

structurefile

toobtain

andeach

subsequenttim

eframe

isfitted

tothe

previouslyfitted

structure.T

hisw

aya

continuoustrajectory

isgenerated,

which

might

notbe

thecase

when

usingthe

regularfit

method,

e.g.w

henyour

proteinundergoes

largeconform

ationaltransitions.

Option

-pb

csets

thetype

ofperiodic

boundarycondition

treatment.w

ho

leputs

theatom

sin

thebox

andthen

makes

brokenm

oleculesw

hole(a

runinput

fileis

required).in

bo

xputs

allthe

atoms

inthe

box.n

oju

mp

checksif

atoms

jump

acrossthe

boxand

thenputs

themback.

This

hasthe

effectthat

allm

oleculesw

illremain

whole

(providedthey

were

whole

inthe

initialconformation),note

thatthisensures

acontinuous

trajectorybutm

oleculesm

aydiffuse

outofthebox.

The

startingconfiguration

forthisprocedure

istaken

fromthe

structurefile,ifone

issupplied,otherw

iseitis

thefirstfram

e.-p

bc

isignored

when-fit

of-pfit

isset,in

thatcasem

oleculesw

illbem

adew

hole.

Option

-ur

setsthe

unitcellrepresentationforoptions

wh

ole

andin

bo

xof-p

bc

.A

llthreeoptions

givedifferent

resultsfor

triclincboxes

andidenticalresults

forrectangular

boxes.re

ctis

theordinary

brickshape.tric

isthe

triclinicunitcell.co

mp

act

putsallatom

satthe

closestdistancefrom

thecenter

ofthebox.

This

canbe

usefulforvisualizing

e.g.truncated

octahedrons.

Option

-cen

ter

centersthe

systemin

thebox.

The

usercan

selectthegroup

which

isused

todeterm

inethe

geometricalcenter.

Use

option-pbc

wh

ole

inaddition

to-ce

nte

rw

henyou

want

allmolecules

inthe

boxafter

thecentering.

With

-dt

itispossible

toreduce

thenum

beroffram

esin

theoutput.

This

optionrelies

onthe

accuracyof

thetim

esin

yourinputtrajectory,so

iftheseare

inaccurateuse

the-tim

este

poption

tom

odifythe

time

(thiscan

bedone

simultaneously).

Page 23: CS USER AL Simulations - GROMACS

22

6A

pp

en

dix

E.

Ma

nu

alP

age

s

E.6

5tp

bcon

v

tpbc

onv

can

edit

run

inpu

tfile

sin

two

way

s.

1st.

bycr

eatin

ga

run

inpu

tfile

for

aco

ntin

uatio

nru

nw

hen

your

sim

ulat

ion

has

cras

hed

due

toe.

g.a

full

disk

,orb

ym

akin

ga

cont

inua

tion

run

inpu

tfile

.N

ote

that

afr

ame

with

coor

dina

tes

and

velo

citie

sis

need

ed,

whi

chm

eans

that

whe

nyo

une

ver

writ

eve

loci

ties,

you

can

not

use

tpbc

onv

and

you

have

tost

art

the

run

agai

nfr

omth

ebe

ginn

ing.

2nd.

bycr

eatin

ga

tpx

file

for

asu

bset

ofyo

uror

igin

altp

xfil

e,w

hich

isus

eful

whe

nyo

uw

antt

ore

mov

eth

eso

lven

tfro

myo

urtp

xfil

e,or

whe

nyo

uw

antt

om

ake

e.g.

apu

reC

atp

xfil

e.W

AR

NIN

G:t

his

tpx

file

isno

tful

lyfu

nctio

nal.

File

s-s

top

ol.t

pr

Inpu

tG

ener

icru

nin

put:

tpr

tpb

tpa

-ftr

aj.t

rrIn

put,

Opt

.F

ullp

reci

sion

traj

ecto

ry:

trr

trj

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e-o

tpxo

ut.tp

rO

utpu

tG

ener

icru

nin

put:

tpr

tpb

tpa

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-t

ime

real

-1C

ontin

uefr

omfr

ame

atth

istim

e(p

s)in

stea

dof

the

last

fram

e-e

xte

nd

real

0E

xten

dru

ntim

eby

this

amou

nt(p

s)-u

ntil

real

0E

xten

dru

ntim

eun

tilth

isen

ding

time

(ps)

-un

con

stra

ine

dbo

olye

sF

ora

cont

inuo

ustr

ajec

tory

,th

eco

nstr

aint

ssh

ould

not

beso

lved

befo

reth

efir

stst

ep(d

efau

lt)

E.6

6tr

jcat

trjc

atco

ncat

enat

esse

vera

linp

uttr

ajec

tory

files

inso

rted

orde

r.In

case

ofdo

uble

time

fram

esth

eon

ein

the

late

rfil

eis

used

.B

ysp

ecify

ing-s

ettim

eyo

uw

illbe

aske

dfo

rth

est

artt

ime

ofea

chfil

e.T

hein

putfi

les

are

take

nfr

omth

eco

mm

and

line,

such

that

aco

mm

and

like

trjc

at

-ofix

ed

.trr

*.tr

rsh

ould

doth

etr

ick.

File

s-o

tra

jou

t.xt

cO

utpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-bre

al-1

Firs

ttim

eto

use

-ere

al-1

Last

time

tous

e-d

tre

al0

Onl

yw

rite

fram

ew

hen

tMO

Ddt

=fir

sttim

e-p

rec

int

3P

reci

sion

for

.xtc

and

.gro

writ

ing

innu

mbe

rof

deci

mal

plac

es-v

el

bool

yes

Rea

dan

dw

rite

velo

citie

sif

poss

ible

-se

ttim

ebo

oln

oC

hang

est

artin

gtim

ein

tera

ctiv

ely

-so

rtbo

olye

sS

ortt

raje

ctor

yfil

es(n

otfr

ames

)

1.2

.M

ole

cula

rD

yna

mic

sS

imu

latio

ns

3

type

ofw

aven

umbe

rty

peof

bond

vibr

atio

n(c

m−

1)

C-H

,O-H

,N-H

stre

tch

3000

–350

0C

=C

,C=

O,

stre

tch

1700

–200

0H

OH

bend

ing

1600

C-C

stre

tch

1400

–160

0H

2C

Xsc

iss,

rock

1000

–150

0C

CC

bend

ing

800–

1000

O-H···O

libra

tion

400–

700

O-H···O

stre

tch

50–

200

Tabl

e1.

1:Ty

pica

lvib

ratio

nalf

requ

enci

es(w

aven

umbe

rs)

inm

olec

ules

and

hydr

ogen

-bon

ded

liq-

uids

.C

ompa

rekT/h

=20

0cm

−1

at30

0K

.

repr

esen

tatr

aje

cto

ryof

the

syst

em.

Afte

rin

itial

chan

ges,

the

syst

emw

illus

ually

reac

han

eq

ui-

libriu

mst

ate.

By

aver

agin

gov

eran

equi

libriu

mtr

ajec

tory

man

ym

acro

scop

icpr

oper

ties

can

beex

trac

ted

from

the

outp

utfil

e.

Itis

usef

ulat

this

poin

tto

cons

ider

the

limita

tions

ofM

Dsi

mul

atio

ns.

The

user

shou

ldbe

awar

eof

thos

elim

itatio

nsan

dal

way

spe

rfor

mch

ecks

onkn

own

expe

rimen

talp

rope

rtie

sto

asse

ssth

eac

cura

cyof

the

sim

ulat

ion.

We

listt

heap

prox

imat

ions

belo

w.

The

sim

ulat

ions

are

clas

sica

lU

sing

New

ton’

seq

uatio

nof

mot

ion

auto

mat

ical

lyim

plie

sth

eus

eof

cla

ssic

alm

ech

an

icsto

desc

ribe

the

mot

ion

ofat

oms.

Thi

sis

allr

ight

for

mos

tat

oms

atno

rmal

tem

pera

ture

s,bu

tth

ere

are

exce

ptio

ns.

Hyd

roge

nat

oms

are

quite

light

and

the

mot

ion

ofpr

oton

sis

som

etim

esof

esse

ntia

lqua

ntum

mec

hani

calc

hara

cter

.F

orex

ampl

e,a

prot

onm

aytu

nn

elth

roug

ha

po-

tent

ialb

arrie

rin

the

cour

seof

atr

ansf

erov

era

hydr

ogen

bond

.S

uch

proc

esse

sca

nnot

bepr

oper

lytr

eate

dby

clas

sica

ldyn

amic

s!H

eliu

mliq

uid

atlo

wte

mpe

ratu

reis

anot

here

xam

ple

whe

recl

assi

calm

echa

nics

brea

ksdo

wn.

Whi

lehe

lium

may

notd

eepl

yco

ncer

nus

,the

high

freq

uenc

yvi

brat

ions

ofco

vale

ntbo

nds

shou

ldm

ake

usw

orry

!T

hest

atis

tical

mec

hani

csof

acl

assi

calh

arm

onic

osci

llato

rdi

ffers

appr

ecia

bly

from

that

ofa

real

quan

tum

osci

llato

r,w

hen

the

reso

nanc

efr

eque

ncyν

appr

oxim

ates

orex

ceed

sk BT/h

.N

owat

room

tem

per-

atur

eth

ew

aven

umbe

rσ=

1/λ

=ν/c

atw

hichhν

=k

BT

isap

prox

imat

ely

200

cm−1.

Thu

sal

lfre

quen

cies

high

erth

an,s

ay,1

00cm−1

are

susp

ecto

fmis

beha

vior

incl

assi

cals

im-

ulat

ions

.T

his

mea

nsth

atpr

actic

ally

allb

ond

and

bond

-ang

levi

brat

ions

are

susp

ect,

and

even

hydr

ogen

-bon

ded

mot

ions

astr

ansl

atio

nalo

rlib

ratio

nalH

-bon

dvi

brat

ions

are

beyo

ndth

ecl

assi

call

imit

(see

Tabl

e1.1)

.W

hatc

anw

edo

?

Wel

l,ap

artf

rom

real

quan

tum

-dyn

amic

alsi

mul

atio

ns,w

eca

ndo

eith

erof

two

thin

gs:

(a)

Ifw

epe

rfor

mM

Dsi

mul

atio

nsus

ing

harm

onic

osci

llato

rsfo

rbo

nds,

we

shou

ldm

ake

corr

ectio

nsto

the

tota

lint

erna

lene

rgyU

=E

kin

+E

potan

dsp

ecifi

che

atCV

(and

toen

trop

yS

and

free

ener

gyAorG

ifth

ose

are

calc

ulat

ed).

The

corr

ectio

nsto

the

ener

gyan

dsp

ecifi

che

atof

aon

e-di

men

sion

alos

cilla

tor

with

freq

uenc

are:

[6]

UQ

M=U

cl+kT

( 1 2x−

1+

x

ex−

1

)(1

.3)

Page 24: CS USER AL Simulations - GROMACS

4C

ha

pte

r1

.In

trod

uctio

n

CQ

MV

=C

clV

+k (

x2e

x

(ex−

1)2−

1 ),

(1.4)

where

x=hν/kT

.T

heclassicaloscillator

absorbstoo

much

energy(

kT

),while

thehigh-

frequencyquantum

oscillatoris

inits

groundstate

atthezero-pointenergy

levelof12 hν.

(b)W

ecan

treatthebonds

(andbond

angles)as

con

strain

tsinthe

equationofm

otion.T

herationalbehind

thisis

thata

quantumoscillator

inits

groundstate

resembles

aconstrained

bondm

oreclosely

thana

classicaloscillator.

Agood

practicalreason

forthis

choiceis

thatthe

algorithmcan

uselarger

time

stepsw

henthe

highestfrequencies

arerem

oved.In

practicethe

time

stepcan

bem

adefourtim

esas

largew

henbonds

areconstrained

thanw

henthey

areoscillators

[7].G

RO

MA

CS

hasthis

optionfor

thebonds,and

forthe

bondangles.

The

flexibilityofthe

latteris

ratheressentialto

allowfor

therealistic

motion

andcoverage

ofconfigurationalspace[

7].

Electrons

arein

theground

stateIn

MD

we

useaco

nse

rvativeforce

fieldthatis

afunction

ofthepositions

ofatoms

only.T

hism

eansthat

theelectronic

motions

arenot

considered:the

electronsare

supposedto

adjusttheir

dynamics

infinitelyfast

when

theatom

icpositions

change(theB

orn

-Op

pe

nh

eim

er

approximation),and

remain

intheir

groundstate.

This

isreally

allright,almostalw

ays.B

utof

course,electron

transferprocesses

andelectronically

excitedstates

cannot

betreated.

Neither

canchem

icalreactionsbe

treatedproperly,

butthereare

otherreasons

toshy

away

fromreactions

forthe

time

being.

Force

fieldsare

approximate

Force

fieldsprovide

theforces.

They

arenotreally

apartofthe

simulation

method

andtheir

parameters

canbe

user-modified

asthe

needarises

orknow

ledgeim

proves.B

utthe

formof

theforces

thatcan

beused

ina

particularprogram

issubject

tolim

itations.T

heforce

fieldthat

isincorporated

inG

RO

MA

CS

isdescribed

inC

hapter4.

Inthe

presentversion

theforce

fieldis

pair-additive(apartfrom

long-rangecoulom

bforces),itcannotincorporate

polarizabilities,and

itdoes

notcontain

fine-tuningof

bondedinteractions.

This

urgesthe

inclusionof

some

limitations

inthis

listbelow

.F

orthe

restit

isquite

usefuland

fairlyreliable

forbio

macro-m

oleculesin

aqueoussolution!

The

forcefield

ispair-additive

This

means

thatallnon

-bo

nd

edforces

resultfromthe

sumofnon-bonded

pairinteractions.

Non

pair-additiveinteractions,

them

ostimportantexam

pleofw

hichis

interactionthrough

atomic

polarizability,are

representedbye

ffective

pa

irp

ote

ntia

ls.O

nlyaverage

nonpair-

additivecontributions

areincorporated.

This

alsom

eansthat

thepair

interactionsare

notpure,i.e.,they

arenotvalid

forisolatedpairs

orforsituationsthatdifferappreciably

fromthe

testsystem

son

which

them

odelsw

ereparam

eterized.In

fact,the

effectivepair

potentialsare

notthat

badin

practice.B

utthe

omission

ofpolarizability

alsom

eansthat

electronsin

atoms

donotprovide

adielectric

constantasthey

should.F

orexam

ple,realliquid

alkaneshave

adielectric

constantofslightlym

orethan

2,which

reducethe

long-rangeelectrostatic

interactionbetw

een(partial)

charges.T

husthe

simulations

willexaggerate

thelong-range

Coulom

bterm

s.Luckily,the

nextitemcom

pensatesthis

effectabit.

Long-rangeinteractions

arecut-off

Inthis

versionG

RO

MA

CS

always

usesa

cut-offradius

forthe

Lennard-Jonesinteractions

E.6

4.

pro

ton

ate

22

5

-ncle

an

.nd

xO

utput,Opt.

Indexfile

-qcle

an

.pd

bO

utput,Opt.

Generic

structure:gro

g96pdb

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int0

Setthe

nicelevel-m

erg

ebool

no

Merge

multiple

chainsinto

onem

olecule-in

ter

booln

oS

etthenext6

optionsto

interactive-ss

booln

oInteractive

SS

bridgeselection

-ter

booln

oInteractive

terminiselection,iso

charged-lys

booln

oInteractive

Lysineselection,iso

charged-a

spbool

no

InteractiveA

sparticA

cidselection,iso

charged-g

lubool

no

InteractiveG

lutamic

Acid

selection,isocharged

-his

booln

oInteractive

Histidine

selection,isochecking

H-bonds

-an

gle

real1

35

Minim

umhydrogen-donor-acceptor

anglefor

aH

-bond(degrees)

-dist

real0

.3M

aximum

donor-acceptordistance

fora

H-bond

(nm)

-un

abool

no

Select

aromatic

ringsw

ithunited

CH

atoms

onP

henylalanine,T

rypto-phane

andTyrosine

-sort

boolye

sS

orttheresidues

accordingto

database-H

14

booln

oU

se1-4

interactionsbetw

eenhydrogen

atoms

-ign

hbool

no

Ignorehydrogen

atoms

thatarein

thepdb

file-a

lldih

booln

oG

enerateallproper

dihedrals-d

um

my

enumn

on

eC

onvertatoms

todum

my

atoms:n

on

e,h

ydro

ge

ns

oraro

ma

tics-h

ea

vyhbool

no

Make

hydrogenatom

sheavy

-de

ute

rate

booln

oC

hangethe

mass

ofhydrogensto

2am

u

E.64

protonate

pro

ton

ate

reads(a)

conformation(s)

andadds

allm

issinghydrogens

asdefined

inffg

mx2

.hd

b.

Ifonly

-sis

specified,thisconform

ationw

illbeprotonated,ifalso

-fis

specified,theconform

ation(s)w

illbe

readfrom

thisfile

which

canbe

eithera

singleconform

ationor

atrajectory.

Ifapdb

fileis

supplied,residuenam

esm

ightnotcorrespondto

tothe

GR

OM

AC

Snam

ingconventions,in

which

casethese

residuesw

illprobablynotbe

properlyprotonated.

Ifanindex

fileis

specified,pleasenote

thattheatom

numbers

shouldcorrespond

tothe

protonatedstate.

Files

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-f

traj.xtc

Input,Opt.

Generic

trajectory:xtc

trrtrjgro

g96pdb

-nin

de

x.nd

xInput,O

pt.Index

file-o

pro

ton

ate

d.xtc

Output

Generic

trajectory:xtc

trrtrjgro

g96pdb

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int0

Setthe

nicelevel-b

time

-1F

irstframe

(ps)to

readfrom

trajectory-e

time

-1Lastfram

e(ps)

toread

fromtrajectory

-dt

time

-1O

nlyuse

frame

when

tMO

Ddt=

firsttime

(ps)

Page 25: CS USER AL Simulations - GROMACS

22

4A

pp

en

dix

E.

Ma

nu

alP

age

s

E.6

3pd

b2gm

x

Thi

spr

ogra

mre

ads

apd

bfil

e,le

tsyo

uch

oose

afo

rcefi

eld,

read

sso

me

data

base

files

,ad

dshy

drog

ens

toth

em

olec

ules

and

gene

rate

sco

ordi

nate

sin

Gro

mac

s(G

rom

os)

form

atan

da

topo

logy

inG

rom

acs

form

at.

The

sefil

esca

nsu

bseq

uent

lybe

proc

esse

dto

gene

rate

aru

nin

putfi

le.

Not

eth

ata

pdb

file

isno

thin

gm

ore

than

afil

efo

rmat

,and

itne

edno

tnec

essa

rily

cont

ain

apr

otei

nst

ruct

ure.

Eve

ryki

ndof

mol

ecul

efo

rw

hich

ther

eis

supp

ort

inth

eda

taba

seca

nbe

conv

erte

d.If

ther

eis

nosu

ppor

tin

the

data

base

,you

can

add

ityo

urse

lf.

The

prog

ram

has

limite

din

telli

genc

e,it

read

sa

num

ber

ofda

taba

sefil

es,

that

allo

wit

tom

ake

spec

ial

bond

s(C

ys-C

ys,

Hem

e-H

is,

etc.

),if

nece

ssar

yth

isca

nbe

done

man

ually

.T

hepr

ogra

mca

npr

ompt

the

user

tose

lect

whi

chki

ndof

LYS

,A

SP,

GLU

,C

YS

orH

ISre

sidu

esh

ew

ants

.F

orLY

Sth

ech

oice

isbe

twee

nLY

S(t

wo

prot

ons

onN

Z)

orLY

SH

(thr

eepr

oton

s,de

faul

t),

for

AS

Pan

dG

LUun

prot

onat

ed(d

efau

lt)or

prot

onat

ed,

for

HIS

the

prot

onca

nbe

eith

eron

ND

1(H

ISA

),on

NE

2(H

ISB

)or

onbo

th(H

ISH

).B

yde

faul

tthe

sese

lect

ions

are

done

auto

mat

ical

ly.

For

His

,th

isis

base

don

anop

timal

hydr

ogen

bond

ing

conf

orm

atio

n.H

ydro

gen

bond

sar

ede

fined

base

don

asi

mpl

ege

omet

riccr

iteriu

m,

spec

ified

byth

em

axim

umhy

drog

en-d

onor

-acc

epto

ran

gle

and

dono

r-ac

cept

ordi

stan

ce,w

hich

are

setb

y-a

ng

lean

d-d

ist

resp

ectiv

ely.

Opt

ion

-me

rge

will

ask

ifyo

uw

antt

om

erge

cons

ecut

ive

chai

nsin

toon

em

olec

ule,

this

can

beus

eful

for

conn

ectin

gch

ains

with

adi

sulfi

debr

igde

.

pdb2

gmx

will

also

chec

kth

eoc

cupa

ncy

field

ofth

epd

bfil

e.If

any

ofth

eoc

cupa

ncci

esar

eno

ton

e,in

dica

ting

that

the

atom

isno

tres

olve

dw

elli

nth

est

ruct

ure,

aw

arni

ngm

essa

geis

issu

ed.

Whe

na

pdb

file

does

noto

rigin

ate

from

anX

-Ray

stru

ctur

ede

term

inat

ion

allo

ccup

ancy

field

sm

aybe

zero

.E

ither

way

,it

isup

toth

eus

erto

verif

yth

eco

rrec

tnes

sof

the

inpu

tdat

a(r

ead

the

artic

le!)

.

Dur

ing

proc

essi

ngth

eat

oms

will

bere

orde

red

acco

rdin

gto

Gro

mac

sco

nven

tions

.W

ith-n

anin

dex

file

can

bege

nera

ted

that

cont

ains

one

grou

pre

orde

red

inth

esa

me

way

.T

his

allo

ws

you

toco

nver

taG

rom

ostr

ajec

tory

and

coor

dina

tefil

eto

Gro

mos

.T

here

ison

elim

itatio

n:re

orde

ring

isdo

neaf

ter

the

hydr

ogen

sar

est

rippe

dfr

omth

ein

puta

ndbe

fore

new

hydr

ogen

sar

ead

ded.

Thi

sm

eans

that

you

shou

ldno

tuse

-ig

nh

.

The

.gro

and

.g9

6fil

efo

rmat

sdo

nots

uppo

rtch

ain

iden

tifier

s.T

here

fore

itis

usef

ulto

ente

ra

pdb

file

nam

eat

the-o

optio

nw

hen

you

wan

tto

conv

erta

mul

ticha

inpd

bfil

e.

-so

rtw

illso

rtal

lres

idue

sac

cord

ing

toth

eor

deri

nth

eda

taba

se,s

omet

imes

this

isne

cess

ary

toge

tcha

rge

grou

psto

geth

er.

-alld

ihw

illge

nera

teal

lpro

per

dihe

dral

sin

stea

dof

only

thos

ew

ithas

few

hydr

ogen

sas

poss

ible

,thi

sis

usef

ulfo

rus

ew

ithth

eC

harm

mfo

rcefi

eld.

The

optio

n-d

um

my

rem

oves

hydr

ogen

and

fast

impr

oper

dihe

dral

mot

ions

.A

ngul

aran

dou

t-of

-pla

nem

otio

nsca

nbe

rem

oved

bych

angi

nghy

drog

ens

into

dum

my

atom

san

dfix

ing

angl

es,

whi

chfix

esth

eir

posi

tion

rela

tive

tone

ighb

orin

gat

oms.

Add

ition

ally

,all

atom

sin

the

arom

atic

rings

ofth

est

anda

rdam

ino

acid

s(i.

e.P

HE

,T

RP,

TY

Ran

dH

IS)

can

beco

nver

ted

into

dum

my

atom

s,el

min

atin

gth

efa

stim

prop

erdi

hedr

alflu

ctua

tions

inth

ese

rings

.N

ote

that

inth

isca

seal

loth

erhy

drog

enat

oms

are

also

conv

erte

dto

dum

my

atom

s.T

hem

ass

ofal

lato

ms

that

are

conv

erte

din

todu

mm

yat

oms,

isad

ded

toth

ehe

avy

atom

s.

Als

osl

owin

gdo

wn

ofdi

hedr

alm

otio

nca

nbe

done

with-h

ea

vyh

done

byin

crea

sing

the

hydr

ogen

-mas

sby

afa

ctor

of4.

Thi

sis

also

done

for

wat

erhy

drog

ens

tosl

owdo

wn

the

rota

tiona

lmot

ion

ofw

ater

.T

hein

crea

sein

mas

sof

the

hydr

ogen

sis

subt

ract

edfr

omth

ebo

nded

(hea

vy)

atom

soth

atth

eto

talm

ass

ofth

esy

stem

rem

ains

the

sam

e.R

efer

ence

Fee

nstr

aet

al.,

J.C

ompu

t.C

hem

.20

,786

(199

9).

File

s-f

eiw

it.p

db

Inpu

tG

ener

icst

ruct

ure:

gro

g96

pdb

tpr

tpb

tpa

-oco

nf.g

roO

utpu

tG

ener

icst

ruct

ure:

gro

g96

pdb

-pto

po

l.to

pO

utpu

tTo

polo

gyfil

e-i

po

sre

.itp

Out

put

Incl

ude

file

for

topo

logy

1.3

.E

ne

rgy

Min

imiz

atio

na

nd

Se

arc

hM

eth

od

s5

and

som

etim

esal

sofo

rC

oulo

mb.

Due

toth

em

inim

um-im

age

conv

entio

n(o

nly

one

imag

eof

each

part

icle

inth

epe

riodi

cbo

unda

ryco

nditi

ons

isco

nsid

ered

for

apa

irin

tera

ctio

n),t

hecu

t-of

fran

geca

nno

texc

eed

half

the

box

size

.T

hati

sst

illpr

etty

big

for

larg

esy

stem

s,an

dtr

oubl

eis

only

expe

cted

for

syst

ems

cont

aini

ngch

arge

dpa

rtic

les.

But

then

real

bad

thin

gsm

ayha

ppen

,lik

eac

cum

ulat

ion

ofch

arge

sat

the

cut-

offb

ound

ary

orve

ryw

rong

ener

gies

!F

orsu

chsy

stem

syo

ush

ould

cons

ider

usin

gon

eof

the

impl

emen

ted

long

-ran

geel

ectr

osta

tical

gorit

hms.

Bou

ndar

yco

nditi

ons

are

unna

tura

lS

ince

syst

emsi

zeis

smal

l(ev

en10

,000

part

icle

sis

smal

l),a

clus

ter

ofpa

rtic

les

will

have

alo

tof

unw

ante

dbo

unda

ryw

ithits

envi

ronm

ent

(vac

uum

).T

his

we

mus

tav

oid

ifw

ew

ish

tosi

mul

ate

abu

lksy

stem

.S

ow

eus

epe

riodi

cbo

unda

ryco

nditi

ons,

toav

oid

real

phas

ebo

unda

ries.

But

liqui

dsar

eno

tcr

ysta

ls,

soso

met

hing

unna

tura

lre

mai

ns.

Thi

site

mis

men

tione

din

the

last

plac

ebe

caus

eit

isth

ele

aste

vilo

fall.

For

larg

esy

stem

sth

eer

rors

are

smal

l,bu

tfor

smal

lsys

tem

sw

itha

loto

fint

erna

lspa

tialc

orre

latio

n,th

epe

riodi

cbo

unda

ries

may

enha

nce

inte

rnal

corr

elat

ion.

Inth

atca

se,b

ewar

ean

dte

stth

ein

fluen

ceof

syst

emsi

ze.

Thi

sis

espe

cial

lyim

port

ant

whe

nus

ing

latti

cesu

ms

for

long

-ran

geel

ectr

osta

tics,

sinc

eth

ese

are

know

nto

som

etim

esin

trod

uce

extr

aor

derin

g.

1.3

Ene

rgy

Min

imiz

atio

nan

dS

earc

hM

etho

ds

As

men

tione

din

sec.1

.1,i

nm

any

case

sen

ergy

min

imiz

atio

nis

requ

ired.

GR

OM

AC

Spr

ovid

esa

sim

ple

form

oflo

cale

nerg

ym

inim

izat

ion,

theste

ep

est

de

sce

nt

met

hod.

The

pote

ntia

lene

rgy

func

tion

ofa

(mac

ro)m

olec

ular

syst

emis

ave

ryco

mpl

exla

ndsc

ape

(or

hyp

er

surf

ace

)in

ala

rge

num

ber

ofdi

men

sion

s.It

has

one

deep

est

poin

t,th

eg

lob

al

min

imu

man

da

very

larg

enu

mbe

rofl

oca

lm

inim

a,w

here

all

deriv

ativ

esof

the

pote

ntia

len

ergy

func

tion

with

resp

ect

toth

eco

ordi

nate

sar

eze

roan

dal

lse

cond

deriv

ativ

esar

eno

nneg

ativ

e.T

hem

atrix

ofse

cond

deriv

ativ

es,

whi

chis

calle

dth

eHe

ssia

nm

atr

ix,ha

sno

nneg

ativ

eei

genv

alue

s;on

lyth

eco

llect

ive

coor

dina

tes

that

corr

espo

ndto

tran

slat

ion

and

rota

tion

(for

anis

olat

edm

olec

ule)

have

zero

eige

nval

ues.

Inbe

twee

nth

elo

calm

inim

ath

ere

are

sad

dle

po

ints,

whe

reth

eH

essi

anm

atrix

has

only

one

nega

tive

eige

nval

ue.

The

sepo

ints

are

the

mou

ntai

npa

sses

thro

ugh

whi

chth

esy

stem

can

mig

rate

from

one

loca

lmin

imum

toan

othe

r.

Kno

wle

dge

ofal

lloc

alm

inim

a,in

clud

ing

the

glob

alon

e,an

dof

alls

addl

epo

ints

wou

lden

able

usto

desc

ribe

the

rele

vant

stru

ctur

esan

dco

nfor

mat

ions

and

thei

rfr

eeen

ergi

es,

asw

ell

asth

edy

nam

ics

ofst

ruct

ural

tran

sitio

ns.

Unf

ortu

nate

ly,t

hedi

men

sion

ality

ofth

eco

nfigu

ratio

nals

pace

and

the

num

ber

oflo

calm

inim

ais

sohi

ghth

atit

isim

poss

ible

tosa

mpl

eth

esp

ace

ata

suffi

cien

tnu

mbe

rof

poin

tsto

obta

ina

com

plet

esu

rvey

.In

part

icul

ar,

nom

inim

izat

ion

met

hod

exis

tsth

atgu

aran

tees

the

dete

rmin

atio

nof

the

glob

alm

inim

um.

How

ever

,gi

ven

ast

artin

gco

nfigu

ratio

n,it

ispo

ssib

leto

find

then

ea

rest

loca

lm

inim

um.

Nea

rest

inth

isco

ntex

tdo

esno

tal

way

sim

ply

near

est

ina

geom

etric

alse

nse

(i.e.

,th

ele

ast

sum

ofsq

uare

coor

dina

tedi

ffere

nces

),bu

tm

eans

the

min

imum

that

can

bere

ache

dby

syst

emat

ical

lym

ovin

gdo

wn

the

stee

pest

loca

lgr

adie

nt.

Fin

ding

this

near

est

loca

lmin

imum

isal

ltha

tG

RO

MA

CS

can

dofo

ryo

u,so

rry!

Ifyo

uw

ant

tofin

dot

her

min

ima

and

hope

todi

scov

erth

egl

obal

min

imum

inth

epr

oces

s,th

ebe

stad

vice

isto

expe

rimen

twith

tem

pera

ture

-cou

pled

MD

:run

your

syst

emat

ahi

ghte

mpe

ratu

refo

ra

whi

lean

d

Page 26: CS USER AL Simulations - GROMACS

6C

ha

pte

r1

.In

trod

uctio

n

thenquench

itslow

lydow

nto

therequired

temperature;

dothis

repeatedly!If

something

asa

melting

orglass

transitiontem

peratureexists,

itis

wise

tostay

forsom

etim

eslightly

belowthat

temperature

andcooldow

nslow

lyaccording

tosom

eclever

scheme,

aprocess

calledsim

ula

ted

an

ne

alin

g.S

inceno

physicaltruthis

required,youcan

useyour

phantasyto

speedup

thisprocess.

One

trickthatoften

works

isto

make

hydrogenatom

sheavier

(mass

10or

so):although

thatwill

slowdow

nthe

otherwise

veryrapid

motions

ofhydrogenatom

s,itwillhardly

influencethe

slower

motions

inthe

systemw

hileenabling

youto

increasethe

time

stepby

afactor

of3or

4.You

canalso

modify

thepotentialenergy

functionduring

thesearch

procedure,e.g.

byrem

ovingbarriers

(remove

dihedralanglefunctions

orreplace

repulsivepotentials

byso

ftco

repotentials[8]),

butalw

aystake

careto

restorethe

correctfunctionsslow

ly.T

hebestsearch

method

thatallows

ratherdrastic

structuralchangesis

toallow

excursionsinto

four-dimensionalspace

[9],butthis

requiressom

eextra

programm

ingbeyond

thestandard

capabilitiesofG

RO

MA

CS

.

Three

possibleenergy

minim

izationm

ethodsare:

•T

hosethat

requireonly

functionevaluations.

Exam

plesare

thesim

plexm

ethodand

itsvariants.

Astep

ism

adeon

thebasis

ofthe

resultsof

previousevaluations.

Ifderivative

information

isavailable,such

methods

areinferior

tothose

thatusethis

information.

•T

hosethat

usederivative

information.

Since

thepartialderivatives

ofthe

potentialenergyw

ithrespect

toall

coordinatesare

known

inM

Dprogram

s(these

areequal

tom

inusthe

forces)this

classofm

ethodsis

verysuitable

asm

odificationofM

Dprogram

s.

•T

hosethat

usesecond

derivativeinform

ationas

well.

These

methods

aresuperior

intheir

convergenceproperties

nearthe

minim

um:

aquadratic

potentialfunctionis

minim

izedin

onestep!

The

problemis

thatforNparticles

a3N×

3N

matrix

mustbe

computed,stored

andinverted.

Apart

fromthe

extraprogram

ming

toobtain

secondderivatives,

form

ostsystem

sof

interestthis

isbeyond

theavailable

capacity.T

hereare

intermediate

methods

buildingup

theH

essianm

atrixon

thefly,

butthey

alsosuffer

fromexcessive

storagere-

quirements.

So

GR

OM

AC

Sw

illshyaw

ayfrom

thisclass

ofmethods.

The

stee

pe

std

esce

ntmethod,

availablein

GR

OM

AC

S,

isof

thesecond

class.It

simply

takesa

stepin

thedirection

ofthe

negativegradient

(hencein

thedirection

ofthe

force),w

ithoutany

considerationofthe

historybuiltup

inprevious

steps.T

hestep

sizeis

adjustedsuch

thatthesearch

isfastbutthe

motion

isalw

aysdow

nhill.T

hisis

asim

pleand

sturdy,butsomew

hatstupid,method:

itsconvergence

canbe

quiteslow

,especially

inthe

vicinityof

thelocal

minim

um!

The

fasterconvergingco

nju

ga

teg

rad

ien

tm

eth

od(see

e.g.[10])

usesgradient

information

fromprevious

steps.In

general,steepestdescentsw

illbringyou

closeto

thenearestlocalm

inimum

veryquickly,

while

conjugategradients

bringsyouveryclose

tothe

localminim

um,butperform

sw

orsefaraw

ayfrom

them

inimum

.

E.6

1.

nm

run

22

3

E.61

nmrun

nmrun

buildsa

Hessian

matrix

fromsingle

conformation.

For

usualNorm

alModes-like

calculations,make

surethat

thestructure

providedis

properlyenergy-m

inimised.

The

generatedm

atrixcan

bediagonalized

byg

nmeig.

Files

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-mh

essia

n.m

txO

utputH

essianm

atrix-g

nm

.log

Output

Logfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-np

int1

Num

berofnodes,m

ustbethe

same

asused

forgrom

pp-v

booln

oVerbose

mode

-com

pa

ctbool

yes

Write

acom

pactlogfile

E.62

options

AllG

RO

MA

CS

programs

have6

standardoptions,ofw

hichsom

eare

hiddenby

default:

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int0

Setthe

nicelevel

•O

ptionalfilesare

notusedunless

theoption

isset,in

contrasttonon

optionalfiles,where

thedefault

filenam

eis

usedw

henthe

optionis

notset.

•A

llGR

OM

AC

Sprogram

sw

illacceptfileoptions

withouta

fileextension

orfilename

beingspecified.

Insuch

casesthe

defaultfilenam

esw

illbe

used.W

ithm

ultipleinput

filetypes,

suchas

genericstructure

format,

thedirectory

will

besearched

forfiles

ofeach

typew

iththe

suppliedor

defaultnam

e.W

henno

suchfile

isfound,or

with

outputfilesthe

firstfiletype

willbe

used.

•A

llGR

OM

AC

Sprogram

sw

iththe

exceptionofmd

run

,n

mru

nand

en

eco

nv

checkif

thecom

-m

andline

optionsare

valid.Ifthis

isnotthe

case,theprogram

willbe

halted.

•E

numerated

options(enum

)shouldbe

usedw

ithone

oftheargum

entslisted

inthe

optiondescription,

theargum

entmay

beabbreviated.

The

firstmatch

tothe

shortestargumentin

thelistw

illbeselected.

•Vector

optionscan

beused

with

1or

3param

eters.W

henonly

oneparam

eteris

suppliedthe

two

othersare

alsosetto

thisvalue.

•F

orm

anyG

RO

MA

CS

programs,the

time

optionscan

besupplied

indifferenttim

eunits,depending

onthe

settingofthe-tu

option.

•A

llGR

OM

AC

Sprogram

scan

readcom

pressedor

g-zippedfiles.

There

might

bea

problemw

ithreading

compressed.xtc

,.trrand

.trjfiles,butthese

willnotcom

pressvery

wellanyw

ay.

•M

ostG

RO

MA

CS

programs

canprocess

atrajectory

with

lessatom

sthan

therun

inputor

structurefile,butonly

ifthetrajectory

consistsofthe

firstnatom

softhe

runinputor

structurefile.

•M

anyG

RO

MA

CS

programs

willaccept

the-tuoption

toset

thetim

eunits

touse

inoutput

files(e.g.

forxmg

rgraphs

orxpm

matrices)

andin

alltime

options.

Page 27: CS USER AL Simulations - GROMACS

22

2A

pp

en

dix

E.

Ma

nu

alP

age

s

-nic

ein

t1

9S

etth

eni

cele

vel

-de

ffn

mst

ring

Set

the

defa

ultfi

lena

me

for

allfi

leop

tions

-np

int

1N

umbe

rof

node

s,m

ustb

eth

esa

me

asus

edfo

rgr

ompp

-vbo

oln

oB

elo

udan

dno

isy

-co

mp

act

bool

yes

Writ

ea

com

pact

log

file

E.5

9m

kan

gndx

mk

angn

dxm

akes

anin

dex

file

for

calc

ulat

ion

ofan

gle

dist

ribut

ions

etc.

Itus

esa

run

inpu

tfile

(.tp

x)

for

the

defin

ition

sof

the

angl

es,d

ihed

rals

etc.

File

s-s

top

ol.t

pr

Inpu

tG

ener

icru

nin

put:

tpr

tpb

tpa

-na

ng

le.n

dx

Out

put

Inde

xfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-t

ype

enum

an

gle

Type

ofan

gle:

an

gle

,g

96

-an

gle

,d

ihe

dra

l,

imp

rop

er

,ry

cka

ert

-be

llem

an

sor

ph

i-p

si

E.6

0ng

mx

ngm

xis

the

Gro

mac

str

ajec

tory

view

er.

Thi

spr

ogra

mre

ads

atr

ajec

tory

file,

aru

nin

put

file

and

anin

dex

file

and

plot

sa

3Dst

ruct

ure

ofyo

urm

olec

ule

onyo

urst

anda

rdX

Win

dow

scre

en.

No

need

for

ahi

ghen

dgr

aphi

csw

orks

tatio

n,it

even

wor

kson

Mon

ochr

ome

scre

ens.

The

follo

win

gfe

atur

esha

vebe

enim

plem

ente

d:3D

view

,rot

atio

n,tr

ansl

atio

nan

dsc

alin

gof

your

mol

ecul

e(s)

,la

bels

onat

oms,

anim

atio

nof

traj

ecto

ries,

hard

copy

inP

ostS

crip

tfor

mat

,use

rde

fined

atom

-filte

rsru

nson

MIT

-X(r

ealX

),op

enw

indo

ws

and

mot

if,us

erfr

iend

lym

enus

,opt

ion

tore

mov

epe

riodi

city

,opt

ion

tosh

owco

mpu

tatio

nalb

ox.

Som

eof

the

mor

eco

mm

onX

com

man

dlin

eop

tions

can

beus

ed:

-bg,

-fg

chan

geco

lors

,-fo

ntfo

ntna

me,

chan

ges

the

font

.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-b

time

-1F

irstf

ram

e(p

s)to

read

from

traj

ecto

ry-e

time

-1La

stfr

ame

(ps)

tore

adfr

omtr

ajec

tory

-dt

time

-1O

nly

use

fram

ew

hen

tMO

Ddt

=fir

sttim

e(p

s)

•B

alls

optio

ndo

esno

twor

k

•S

ome

times

dum

psco

rew

ithou

tago

odre

ason

Cha

pter

2

Defi

nitio

nsan

dU

nits

2.1

Not

atio

n

The

follo

win

gco

nven

tions

for

mat

hem

atic

alty

pese

tting

are

used

thro

ugho

utth

isdo

cum

ent:

Item

Not

atio

nE

xam

ple

Vect

orB

old

italic

ri

Vect

orLe

ngth

Italic

r i

We

defin

eth

elow

erc

ase

subs

crip

tsi,j,k

andl

tode

note

part

icle

s:ri

isth

ep

osi

tion

vect

oro

fpa

rtic

lei,

and

usin

gth

isno

tatio

n:

rij

=r

j−

ri

(2.1

)

r ij

=|r

ij|

(2.2

)

The

forc

eon

part

iclei

isde

note

dby

Fi

and

Fij

=fo

rce

oni

exer

ted

byj

(2.3

)

Ple

ase

note

that

we

chan

ged

nota

tion

asof

ver.

2.0

tor

ij=

rj−

ri

sinc

eth

isis

the

nota

tion

com

mon

lyus

ed.

Ifyo

uen

coun

ter

aner

ror,

letu

skn

ow.

2.2

MD

units

GR

OM

AC

Sus

esa

cons

iste

ntse

tof

units

that

prod

uce

valu

esin

the

vici

nity

ofun

ityfo

rm

ost

rele

vant

mol

ecul

arqu

antit

ies.

Let

usca

llth

emMD

un

its.

The

basi

cun

itsin

this

syst

emar

enm

,ps

,K,e

lect

ron

char

ge(e

)an

dat

omic

mas

sun

it(u

),se

eTa

ble

2.1.

Con

sist

entw

ithth

ese

units

are

ase

tofd

eriv

edun

its,g

iven

inTa

ble

2.2.

The

elec

tric

conv

ersi

onfa

ctorf

=1

4πε o

=13

8.93

548

5(9)

kJm

ol−

1nm

e−2.

Itre

late

sth

em

echa

nica

lqua

ntiti

esto

the

elec

tric

alqu

antit

ies

asin

V=fq2 r

orF

=fq2 r2

(2.4

)

Page 28: CS USER AL Simulations - GROMACS

8C

ha

pte

r2

.D

efin

ition

sa

nd

Un

its

Quantity

Sym

bolU

nitlength

rnm

=10−

9m

mass

mu

(atomic

mass

unit)=1.6605402(10)×10

−27

kg(1/12

ofthem

assofa

Catom

)1.6605402(10)×

10−

27

kgtim

et

ps=10−

12

scharge

qe

=electronic

charge=1.60217733(49)×

10−

19

Ctem

peratureT

K

Table2.1:

Basic

unitsused

inG

RO

MA

CS

.Num

bersin

parenthesesgive

accuracy.

Quantity

Sym

bolU

nitenergy

E,V

kJm

ol −1

Force

FkJ

mol −

1nm

−1

pressurep

kJm

ol −1

nm−

3=

1030/N

AV

Pa

1.660

54×

106

Pa=

16.6054B

arvelocity

vnm

ps −1

=1000

m/s

dipolem

oment

µe

nmelectric

potentialΦ

kJm

ol −1

e −1

=0.010

364272(3)

Voltelectric

fieldE

kJm

ol −1

nm−

1e −

1=

1.036427

2(3)×10

7V

/m

Table2.2:

Derived

units

Electric

potentialsΦand

electricfieldsE

areinterm

ediatequantities

inthe

calculationofenergies

andforces.

They

donotoccur

insideG

RO

MA

CS

.Iftheyare

usedin

evaluations,thereis

achoice

ofequationsand

relatedunits.

We

recomm

endstrongly

tofollow

theusualpractice

toinclude

thefactorf

inexpressions

thatevaluateΦandE

:

Φ(r)

=f ∑

j

qj

|r−

rj |

(2.5)

E(r)

=f ∑

j

qj(r−

rj )

|r−

rj | 3

(2.6)

With

thesedefinitionsqΦ

isan

energyandqE

isa

force.T

heunits

arethose

givenin

Table2.2:

about10m

Vfor

potential.T

husthe

potentialofanelectronic

chargeata

distanceof1

nmequals

f≈

140units≈

1.4V.(exactvalue:

1.439965V

)

Note

thattheseunits

arem

utuallyconsistent;changing

anyofthe

unitsis

likelyto

produceincon-

sistenciesand

isthereforestro

ng

lyd

iscou

raged!In

particular:ifA

areused

insteadofnm

,theunit

oftim

echanges

to0.1

ps.If

thekcal/m

ol(=4.184

kJ/mol)

isused

insteadof

kJ/molfor

energy,the

unitoftime

becomes

0.488882ps

andthe

unitoftemperature

changesto

4.184K

.Butin

bothcases

allelectricalenergiesgo

wrong,because

theyw

illstillbecom

putedin

kJ/mol,expecting

nmas

theunitoflength.

Although

carefulrescalingofcharges

may

stillyieldconsistency,

itisclear

thatsuchconfusions

mustbe

rigidlyavoided.

Interm

softhe

MD

unitsthe

usualphysicalconstantstake

ondifferentvalues,

seeTable

2.3.A

llquantities

areper

mol

ratherthan

perm

olecule.T

hereis

nodistinction

between

Boltzm

ann’sconstantk

andthe

gasconstant

R:

theirvalue

is0.008314

51kJ

mol −

1K−

1.

E.5

8.

md

run

22

1

E.58

mdrun

The

mdrun

programperform

sM

olecularD

ynamics

simulations.

Itreads

therun

inputfile

(-s

)and

dis-tributes

thetopology

overnodes

ifneeded.

The

coordinatesare

passedaround,

sothat

computations

canbegin.

Firsta

neighborlistism

ade,thenthe

forcesare

computed.

The

forcesare

globallysum

med,and

thevelocities

andpositions

areupdated.

Ifnecessaryshake

isperform

edto

constrainbond

lengthsand/or

bondangles.

Temperature

andP

ressurecan

becontrolled

usingw

eakcoupling

toa

bath.

mdrun

producesatleastthree

outputfile,plusone

logfile

(-g

)per

node.T

hetrajectory

file(

-o),contains

coordinates,velocitiesand

optionallyforces.

The

structurefile

(-c

)contains

thecoordinates

andvelocities

ofthelaststep.

The

energyfile

(-e

)contains

energies,thetem

perature,pressure,etc,alotofthese

thingsare

alsoprinted

inthe

logfile

ofnode0.

Optionally

coordinatescan

bew

rittento

acom

pressedtrajectory

file(-x

).

When

runningin

parallelwith

PV

Mor

anold

versionofM

PIthe

-np

optionm

ustbegiven

toindicate

thenum

berofnodes.

The

option-dg

dl

isonly

usedw

henfree

energyperturbation

isturned

on.

With

-reru

nan

inputtrajectorycan

begiven

forw

hichforces

andenergies

willbe

(re)calculated.N

eigh-bor

searchingw

illbeperform

edfor

everyfram

e,unlessn

stlistis

zero(see

the.md

pfile).

ED

(essentialdynamics)

sampling

issw

itchedon

byusing

the-e

iflag

followed

byan.e

di

file.T

he.e

di

filecan

beproduced

usingoptions

inthe

essdynm

enuofthe

WH

ATIF

program.

mdrun

producesa

.ed

ofile

thatcontainsprojections

ofpositions,velocitiesand

forcesonto

selectedeigenvectors.

The

-tableoption

canbe

usedto

passm

druna

formatted

tablew

ithuser-defined

potentialfunctions.T

hefile

isread

fromeither

thecurrentdirectory

orfrom

theG

MX

LIBdirectory.

Anum

berofpreform

attedtables

arepresented

inthe

GM

XLIB

dir,for

6-8,6-9,

6-10,6-11,

6-12Lennard

Jonespotentials

with

normal

Coulom

b.

The

options-pi

,-po

,-pd

,-pn

areused

forpotentialofm

eanforce

calculationsand

umbrella

sampling.

See

manual.

When

mdrun

receivesa

TE

RM

signal,itwillsetnsteps

tothe

currentstepplus

one.W

henm

drunreceives

aU

SR

1signal,

itw

illset

nstepsto

thenext

multiple

ofnstxout

afterthe

currentstep.

Inboth

casesall

theusualoutputw

illbew

rittento

file.W

henrunning

with

MP

I,asignalto

oneofthe

mdrun

processesis

sufficient,thissignalshould

notbesentto

mpirun

orthe

mdrun

processthatis

theparentofthe

others.

Files

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-otra

j.trrO

utputF

ullprecisiontrajectory:

trrtrj

-xtra

j.xtcO

utput,Opt.

Com

pressedtrajectory

(portablexdr

format)

-cco

nfo

ut.g

roO

utputG

enericstructure:

grog96

pdb-e

en

er.e

dr

Output

Generic

energy:edr

ene-g

md

.log

Output

Logfile

-dg

dl

dg

dl.xvg

Output,O

pt.xvgr/xm

grfile

-tab

leta

ble

.xvgInput,O

pt.xvgr/xm

grfile

-reru

nre

run

.xtcInput,O

pt.G

enerictrajectory:

xtctrr

trjgrog96

pdb-e

isa

m.e

di

Input,Opt.

ED

sampling

input-e

osa

m.e

do

Output,O

pt.E

Dsam

plingoutput

-pi

pu

ll.pp

aInput,O

pt.P

ullparameters

-po

pu

llou

t.pp

aO

utput,Opt.

Pullparam

eters-p

dp

ull.p

do

Output,O

pt.P

ulldataoutput

-pn

pu

ll.nd

xInput,O

pt.Index

file

Other

options-h

booln

oP

rinthelpinfo

andquit

Page 29: CS USER AL Simulations - GROMACS

22

0A

pp

en

dix

E.

Ma

nu

alP

age

s

-tim

ere

al-1

Take

fram

eat

orfir

staf

ter

this

time.

-np

int

1G

ener

ate

stat

usfil

efo

r#

node

s-s

hu

ffle

bool

no

Shu

ffle

mol

ecul

esov

erno

des

-so

rtbo

oln

oS

ortm

olec

ules

acco

rdin

gto

Xco

ordi

nate

-rm

du

mb

ds

bool

yes

Rem

ove

cons

tant

bond

edin

tera

ctio

nsw

ithdu

mm

ies

-lo

ad

strin

gR

elea

tive

load

capa

city

ofea

chno

deon

apa

ralle

lmac

hine

.B

esu

reto

use

quot

esar

ound

the

strin

g,w

hich

shou

ldco

ntai

na

num

ber

for

each

node

-ma

xwa

rnin

t1

0N

umbe

rof

war

ning

saf

ter

whi

chin

putp

roce

ssin

gst

ops

-ch

eck

14

bool

no

Rem

ove

1-4

inte

ract

ions

with

outV

ande

rW

aals

E.5

6hi

ghw

ay

high

way

isth

egr

omac

shi

ghw

aysi

mul

ator

.It

isan

X-w

indo

ws

gadg

etth

atsh

ows

a(p

erio

dic)

auto

bahn

with

aus

erde

fined

num

ber

ofca

rs.

Fog

can

betu

rned

onor

offt

oin

crea

seth

enu

mbe

rof

cras

hes.

Nic

efo

ra

back

grou

ndC

PU

-eat

er

File

s-f

hig

hw

ay.

da

tIn

put

Gen

eric

data

file

-aa

uto

.da

tIn

put

Gen

eric

data

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-b

time

-1F

irstf

ram

e(p

s)to

read

from

traj

ecto

ry-e

time

-1La

stfr

ame

(ps)

tore

adfr

omtr

ajec

tory

-dt

time

-1O

nly

use

fram

ew

hen

tMO

Ddt

=fir

sttim

e(p

s)

E.5

7m

ake

ndx

Inde

xgr

oups

are

nece

ssar

yfo

ral

mos

tev

ery

grom

acs

prog

ram

.A

llth

ese

prog

ram

sca

nge

nera

tede

faul

tin

dex

grou

ps.

You

ON

LYha

veto

use

mak

endx

whe

nyo

une

edS

PE

CIA

Lin

dex

grou

ps.

The

reis

ade

faul

tin

dex

grou

pfo

rth

ew

hole

syst

em,9

defa

ulti

ndex

grou

psar

ege

nera

ted

for

prot

eins

,ade

faul

tind

exgr

oup

isge

nera

ted

for

ever

yot

her

resi

due

nam

e.

Whe

nno

inde

xfil

eis

supp

lied,

also

mak

endx

will

gene

rate

the

defa

ultg

roup

s.W

ithth

ein

dex

edito

ryo

uca

nse

lect

onat

om,

resi

due

and

chai

nna

mes

and

num

bers

,yo

uca

nus

eN

OT,

AN

Dan

dO

R,y

ouca

nsp

litgr

oups

into

chai

ns,r

esid

ues

orat

oms.

You

can

dele

tean

dre

nam

egr

oups

.

The

atom

num

berin

gin

the

edito

ran

dth

ein

dex

file

star

tsat

1.

File

s-f

con

f.g

roIn

put

Gen

eric

stru

ctur

e:gr

og9

6pd

btp

rtp

btp

a-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-oin

de

x.n

dx

Out

put

Inde

xfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l

2.3

.R

ed

uce

du

nits

9

Sym

bol

Nam

eVa

lue

NA

VA

voga

dro’

snu

mbe

r6.

022

1367(

36)×

1023

mol−

1

Rga

sco

nsta

nt8.

314

510(

70)×

10−

3kJ

mol−

1K−

1

kB

Bol

tzm

ann’

sco

nsta

ntid

emh

Pla

nck’

sco

nsta

nt0.

399

03132

(24)

kJm

ol−

1ps

hD

irac’

sco

nsta

nt0.

063

50780

7(38

)kJ

mol−

1ps

cve

loci

tyof

light

299

792.

458

nm/p

s

Tabl

e2.

3:S

ome

Phy

sica

lCon

stan

ts

Qua

ntity

Sym

bol

Rel

atio

nto

SI

Leng

thr∗

rσ−

1

Mas

sm∗

mM−

1

Tim

et∗

tσ−

1√ ε/

MTe

mpe

ratu

reT∗

k BTε−

1

Ene

rgy

E∗

Eε−

1

For

ceF∗

Fσε−

1

Pre

ssur

eP∗

3ε−

1

Velo

city

v∗v√ M

/εD

ensi

tyρ∗

3V−

1

Tabl

e2.

4:R

educ

edLe

nnar

d-Jo

nes

quan

titie

s

2.3

Red

uced

units

Whe

nsi

mul

atin

gLe

nnar

d-Jo

nes

(LJ)

syst

ems

itm

ight

bead

vant

ageo

usto

use

redu

ced

units

(i.e

.,se

ttingε i

i=σ

ii=m

i=k

B=

1fo

ron

ety

peof

atom

s).

Thi

sis

poss

ible

.W

hen

spec

ifyin

gth

ein

put

inre

duce

dun

its,

the

outp

utw

illal

sobe

inre

duce

dun

its.

The

reis

one

exce

ptio

n:th

ete

mp

era

ture,

whi

chis

expr

esse

din0.0

0831

451

redu

ced

units

.T

his

isa

cons

eque

nce

ofth

eus

eof

Bol

tzm

ann’

sco

nsta

ntin

the

eval

uatio

nof

tem

pera

ture

inth

eco

de.

Thu

sno

tT

,bu

tkBT

isth

ere

duce

dte

mpe

ratu

re.

AG

RO

MA

CS

tem

pera

ture

T=

1m

eans

are

duce

dte

mpe

ratu

reof

0.00

8...

units

;if

are

duce

dte

mpe

ratu

reof

1is

requ

ired,

the

GR

OM

AC

Ste

mpe

ratu

resh

ould

be12

0.27

17.

InTa

ble

2.4

quan

titie

sar

egi

ven

for

LJpo

tent

ials

:

VL

J=

[ ( σ r

) 12−( σ r

) 6](2

.7)

Page 30: CS USER AL Simulations - GROMACS

10

Ch

ap

ter

2.

De

finitio

ns

an

dU

nits

E.5

5.

gro

mp

p2

19

grompp

callsthe

c-preprocessorto

resolveincludes,

macros

etcetera.To

specifya

macro-preprocessor

otherthan

/lib/cpp(such

asm

4)you

canput

aline

inyour

parameter

filespecifying

thepath

tothat

cpp.S

pecifying-pp

willgetthe

pre-processedtopology

filew

rittenout.

Ifyoursystem

doesnothave

ac-preprocessor,you

canstilluse

grompp,butyou

donothave

accessto

thefeatures

fromthe

cpp.C

omm

andline

optionsto

thec-preprocessor

canbe

givenin

the.m

dp

file.S

eeyour

localmanual(m

ancpp).

When

usingposition

restraintsa

filew

ithrestraint

coordinatescan

besupplied

with

-r,

otherwise

con-straining

willbe

donerelative

tothe

conformation

fromthe-c

option.

Starting

coordinatescan

beread

fromtrajectory

with

-t.

The

lastfram

ew

ithcoordinates

andvelocities

willbe

read,unless

the-time

optionis

used.N

otethatthese

velocitiesw

illnotbeused

when

ge

nve

l=

yes

inyour.m

dp

file.Ifyou

wantto

continuea

crashedrun,itis

easierto

usetp

bco

nv

.

When

preparingan

inputfilefor

parallelm

dru

nitm

aybe

advantageousto

partitionthe

simulation

systemover

thenodes

ina

way

inw

hicheach

nodehas

asim

ilaram

ountof

work.

The

-shuffleoption

doesjust

that.F

ora

singleprotein

inw

aterthis

doesnot

make

adifference,

however

fora

systemw

hereyou

havem

anycopies

ofdifferentmolecules

(e.g.liquid

mixture

orm

embrane/w

atersystem

)the

optionis

definitelya

must.

Afurther

optimization

forparallelsystem

sis

the-sort

optionw

hichsorts

molecules

accordingto

coor-dinates.

This

mustalw

aysbe

usedin

conjunctionw

ith-sh

uffle

,however

sortingalso

works

when

youhave

onlyone

molecule

type.

Using

the-mo

rseoption

grompp

canconvert

theharm

onicbonds

inyour

topologyto

morse

potentials.T

hism

akesit

possibleto

breakbonds.

For

thisoption

tow

orkyou

needan

extrafile

inyour

$GM

XLIB

with

dissociationenergy.

Use

the-debug

optionto

getm

oreinform

ationon

thew

orkingsof

thisoption

(lookfor

MO

RS

Ein

thegrom

pp.logfile

usingless

orsom

ethinglike

that).

By

defaultallbonded

interactionsw

hichhave

constantenergy

dueto

dumm

yatom

constructionsw

illberem

oved.If

thisconstant

energyis

notzero,

thisw

illresult

ina

shiftin

thetotal

energy.A

llbonded

interactionscan

bekeptby

turningoff-rm

du

mb

ds

.A

dditionally,allconstraintsfor

distancesw

hichw

illbe

constantanyw

aybecause

ofdum

my

atomconstructions

will

berem

oved.If

anyconstraints

remain

which

involvedum

my

atoms,a

fatalerrorw

illresult.

Toverify

yourrun

inputfile,pleasem

akenotice

ofallwarnings

onthe

screen,andcorrectw

herenecessary.

Do

alsolook

atthecontents

ofthemdo

ut.m

dp

file,thiscontains

comm

entlines,asw

ellasthe

inputthatg

rom

pp

hasread.

Ifin

doubtyou

canstart

grompp

with

the-d

eb

ug

optionw

hichw

illgiveyou

more

information

ina

filecalled

grompp.log

(alongw

ithrealdebug

info).F

inally,youcan

seethe

contentsofthe

runinputfile

with

thegm

xdu

mp

program.

Files

-fg

rom

pp

.md

pInput

grompp

inputfilew

ithM

Dparam

eters-p

om

do

ut.m

dp

Output

grompp

inputfilew

ithM

Dparam

eters-c

con

f.gro

InputG

enericstructure:

grog96

pdbtpr

tpbtpa

-rco

nf.g

roInput,O

pt.G

enericstructure:

grog96

pdbtpr

tpbtpa

-nin

de

x.nd

xInput,O

pt.Index

file-d

esh

uf

de

shu

f.nd

xO

utput,Opt.

Indexfile

-pto

po

l.top

InputTopology

file-p

pp

roce

ssed

.top

Output,O

pt.Topology

file-o

top

ol.tp

rO

utputG

enericrun

input:tpr

tpbtpa

-ttra

j.trrInput,O

pt.F

ullprecisiontrajectory:

trrtrj

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int0

Setthe

nicelevel-v

boolye

sB

eloud

andnoisy

Page 31: CS USER AL Simulations - GROMACS

21

8A

pp

en

dix

E.

Ma

nu

alP

age

s

File

s-f

tra

j.xtc

Inpu

t,O

pt.

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b-f

2tr

aj.x

tcIn

put,

Opt

.G

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-s1

top

1.tp

rIn

put,

Opt

.G

ener

icru

nin

put:

tpr

tpb

tpa

-s2

top

2.tp

rIn

put,

Opt

.G

ener

icru

nin

put:

tpr

tpb

tpa

-cto

po

l.tp

rIn

put,

Opt

.S

truc

ture

+m

ass(

db):

tpr

tpb

tpa

gro

g96

pdb

-ee

ne

r.e

dr

Inpu

t,O

pt.

Gen

eric

ener

gy:

edr

ene

-e2

en

er2

.ed

rIn

put,

Opt

.G

ener

icen

ergy

:ed

ren

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-v

dw

fac

real

0.8

Fra

ctio

nof

sum

ofV

dWra

diiu

sed

asw

arni

ngcu

toff

-bo

nlo

real

0.4

Min

.fr

act.

ofsu

mof

VdW

radi

ifor

bond

edat

oms

-bo

nh

ire

al0

.7M

ax.

frac

t.of

sum

ofV

dWra

diif

orbo

nded

atom

s-t

ol

real

0To

lera

nce

for

com

parin

gre

alva

lues

E.5

4gm

xdum

p

gmxd

ump

read

sa

run

inpu

tfil

e(

.tp

a/.t

pr

/.tp

b),

atr

ajec

tory

(.trj

/.trr

/.xtc

)or

anen

ergy

file

(.e

ne

/.ed

r)a

ndpr

ints

that

tost

anda

rdou

tput

ina

read

able

form

at.

Thi

spr

ogra

mis

esse

ntia

lfor

chec

king

your

run

inpu

tfile

inca

seof

prob

lem

s.

File

s-s

top

ol.t

pr

Inpu

t,O

pt.

Gen

eric

run

inpu

t:tp

rtp

btp

a-f

tra

j.xtc

Inpu

t,O

pt.

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b-e

en

er.

ed

rIn

put,

Opt

.G

ener

icen

ergy

:ed

ren

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-n

rbo

olye

sS

how

inde

xnu

mbe

rsin

outp

ut(le

avin

gth

emou

tmak

esco

mpa

rison

eas-

ier,

butc

reat

esa

usel

ess

topo

logy

)

E.5

5gr

ompp

The

grom

acs

prep

roce

ssor

read

sa

mol

ecul

arto

polo

gyfil

e,ch

ecks

the

valid

ityof

the

file,

expa

nds

the

topo

logy

from

am

olec

ular

desc

riptio

nto

anat

omic

desc

riptio

n.T

heto

polo

gyfil

eco

ntai

nsin

form

atio

nab

out

mol

ecul

ety

pes

and

the

num

ber

ofm

olec

ules

,th

epr

epro

cess

orco

pies

each

mol

ecul

eas

need

ed.

The

reis

nolim

itatio

non

the

num

ber

ofm

olec

ule

type

s.B

onds

and

bond

-ang

les

can

beco

nver

ted

into

cons

trai

nts,

sepa

rate

lyfo

rhy

drog

ens

and

heav

yat

oms.

The

na

coor

dina

tefil

eis

read

and

velo

citie

sca

nbe

gene

rate

dfr

oma

Max

wel

lian

dist

ribut

ion

ifre

ques

ted.

grom

ppal

sore

ads

para

met

ers

for

the

mdr

un(e

g.nu

mbe

rof

MD

step

s,tim

est

ep,c

ut-o

ff),a

ndot

hers

such

asN

EM

Dpa

ram

eter

s,w

hich

are

corr

ecte

dso

that

the

neta

ccel

erat

ion

isze

ro.

Eve

ntua

llya

bina

ryfil

eis

prod

uced

that

can

serv

eas

the

sole

inpu

tfile

for

the

MD

prog

ram

.

grom

ppus

esth

eat

omna

mes

from

the

topo

logy

file.

The

atom

nam

esin

the

coor

dina

tefil

e(o

ptio

n-c

)ar

eon

lyre

adto

gene

rate

war

ning

sw

hen

they

dono

tm

atch

the

atom

nam

esin

the

topo

logy

.N

ote

that

the

atom

nam

esar

eirr

elev

antf

orth

esi

mul

atio

nas

only

the

atom

type

sar

eus

edfo

rge

nera

ting

inte

ract

ion

para

met

ers.

Cha

pter

3

Alg

orith

ms

3.1

Intro

duct

ion

Inth

isch

apte

rw

efir

stgi

vede

scrib

etw

oge

nera

lcon

cept

sus

edin

GR

OM

AC

S:

pe

rio

dic

bo

un

da

ryco

nd

itio

ns(

sec.

3.2)

and

theg

rou

pco

nce

pt(s

ec.3

.3).

The

MD

algo

rithm

isde

scrib

edin

sec.3.

4:fir

sta

glob

alfo

rmof

the

algo

rithm

isgi

ven,

whi

chis

refin

edin

subs

eque

ntsu

bsec

tions

.T

he(s

impl

e)E

M(E

nerg

yM

inim

izat

ion)

algo

rithm

isde

scrib

edin

sec.

3.10

.S

ome

othe

ral

gorit

hms

for

spec

ialp

urpo

sedy

nam

ics

are

desc

ribed

afte

rth

is.

Inth

efin

alse

c.3.

14of

this

chap

ter

afe

wpr

inci

ples

are

give

non

whi

chpa

ralle

lizat

ion

ofG

RO

MA

CS

isba

sed.

The

para

lleliz

atio

nis

hard

lyvi

sibl

efo

rth

eus

eran

dis

ther

efor

eno

ttre

ated

inde

tail.

Afe

wis

sues

are

ofge

nera

lin

tere

st.

Inal

lca

ses

the

syst

em

mus

tbe

defin

ed,

cons

istin

gof

mol

ecul

es.

Mol

ecul

esag

ain

cons

ist

ofpa

rtic

les

with

defin

edin

tera

ctio

nfu

nctio

ns.

The

deta

iled

desc

riptio

nof

theto

po

log

yoft

hem

olec

ules

and

ofth

eforc

efie

ldan

dth

eca

lcul

atio

nof

forc

esis

give

nin

chap

ter4.

Inth

epr

esen

tcha

pter

we

desc

ribe

othe

ras

pect

sof

the

algo

rithm

,suc

has

pair

listg

ener

atio

n,up

date

ofve

loci

ties

and

posi

tions

,co

uplin

gto

exte

rnal

tem

pera

ture

and

pres

sure

,co

nser

vatio

nof

cons

trai

nts.

Thea

na

lysi

soft

heda

tage

nera

ted

byan

MD

sim

ulat

ion

istr

eate

din

chap

ter8

.

3.2

Per

iodi

cbo

unda

ryco

nditi

ons

The

clas

sica

lway

tom

inim

ize

edge

effe

cts

ina

finite

syst

emis

toap

ply

pe

rio

dic

bo

un

da

ryco

nd

i-tio

ns.

The

atom

sof

the

syst

emto

besi

mul

ated

are

puti

nto

asp

ace-

fillin

gbo

x,w

hich

issu

rrou

nded

bytr

ansl

ated

copi

esof

itsel

f(F

ig.3.1)

.T

hus

ther

ear

eno

boun

darie

sof

the

syst

em;

the

artif

act

caus

edby

unw

ante

dbo

unda

ries

inan

isol

ated

clus

ter

isno

wre

plac

edby

the

artif

act

ofpe

riodi

cco

nditi

ons.

Ifa

crys

tali

ssi

mul

ated

,su

chbo

unda

ryco

nditi

ons

are

desi

red

(alth

ough

mot

ions

are

natu

rally

rest

ricte

dto

perio

dic

mot

ions

with

wav

elen

gths

fittin

gin

toth

ebo

x).

Ifon

ew

ishe

sto

sim

ulat

eno

n-pe

riodi

csy

stem

s,as

liqui

dsor

solu

tions

,the

perio

dici

tyby

itsel

fcau

ses

erro

rs.

The

erro

rsca

nbe

eval

uate

dby

com

parin

gva

rious

syst

emsi

zes;

they

are

expe

cted

tobe

less

seve

reth

anth

eer

rors

resu

lting

from

anun

natu

ralb

ound

ary

with

vacu

um.

Page 32: CS USER AL Simulations - GROMACS

12

Ch

ap

ter

3.

Alg

orith

ms

j’j’

i’i’

i’i’

j’

i’i’

y

x

y

x

j’j’

i’

i’

i’i

j’

j’j’

j’

i’i

i’

j’j’

j’

j

i’i’

i’

j’

i’i’

j’

j’j’

j

Figure

3.1:P

eriodicboundary

conditionsin

two

dimensions.

There

areseveralpossible

shapesfor

space-fillingunitcells.

Som

e,asthe

rho

mb

icd

od

eca

he

dro

nand

thetrun

cate

do

ctah

ed

ron[11]

approacha

sphericalshape

betterthan

acubic

boxand

aretherefore

more

economicalfor

studyingan

(approximately

spherical)m

acromolecule

insolution,

sinceless

solventmolecules

arerequired

tofillthe

boxgiven

am

inimum

distancebetw

eenm

acro-m

olecularim

ages.H

owever,

aperiodic

systembased

onthe

rhombic

dodecahedronor

truncatedoctahedron

isequivalentto

aperiodic

systembased

ona

triclinic

unitcell.T

helatter

shapeis

them

ostgeneralspace-filling

unitcell;

itcom

prisesallpossible

space-fillingshapes

[12].

Therefore

GR

OM

AC

Sis

basedon

thetriclinic

unitcell.

GR

OM

AC

Suses

periodicboundary

conditions,com

binedw

iththe

min

imu

mim

ageco

nven

tion

:only

one-

thenearest

-im

ageof

eachparticle

isconsidered

forshort-range

non-bondedinter-

actionterm

s.F

orlong-range

electrostaticinteractions

thisis

notalw

aysaccurate

enough,and

GR

OM

AC

Stherefore

alsoincorporates

latticesum

methods

likeE

wald

Sum

,PM

Eand

PP

PM

.

Grom

acssupports

triclinicboxes

ofanyshape.

The

boxis

definedby

the3

boxvectorsa

,bandc.

The

boxvectors

mustsatisfy

thefollow

ingconditions:

ay

=a

z=bz

=0

(3.1)

ax>

0,

by>

0,

cz>

0(3.2)

|bx |≤

12a

x ,|c

x |≤12a

x ,|c

y |≤12by

(3.3)

Equations

(3.1)can

always

bestatisfied

byrotating

thebox.

Equations

(3.2)

and(3.3)

canalw

aysbe

statisfiedby

addingand

subtractingbox

vectors.

Even

when

simulating

usinga

triclinicbox,G

RO

MA

CS

always

putsthe

particlesin

abrick

shapedvolum

e,forefficiency

reasons.T

hisis

illustratedin

Fig.

3.1for

a2-dim

ensionalsystem.

So

from

E.5

2.

gen

pr

21

7

Files

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-nin

de

x.nd

xInput,O

pt.Index

file-o

ou

t.gro

Output

Generic

structure:gro

g96pdb

-gg

en

ion

.log

Output

Logfile

-po

tp

ot.p

db

Output,O

pt.P

roteindata

bankfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-np

int0

Num

berofpositive

ions-p

na

me

stringN

aN

ame

ofthepositive

ion-p

qreal

1C

hargeofthe

positiveion

-nn

int0

Num

berofnegative

ions-n

na

me

stringC

lN

ame

ofthenegative

ion-n

qreal

-1C

hargeofthe

negativeion

-rmin

real0

.6M

inimum

distancebetw

eenions

-ran

do

mbool

no

Use

randomplacem

entof

ionsinstead

ofbased

onpotential.

The

rmin

optionshould

stillwork

-see

dint

19

93

Seed

forrandom

number

generator

E.52

genpr

genprproduces

aninclude

filefor

atopology

containinga

listof

atomnum

bersand

threeforce

constantsfor

theX

,Yand

Zdirection.

Asingle

isotropicforce

constantmay

begiven

onthe

comm

andline

insteadofthree

components.

WA

RN

ING

:genpronly

works

forthe

firstmolecule.

Position

restraintsare

interactionsw

ithinm

olecules,therefore

theyshould

beincluded

within

thecorrect

[m

ole

cule

type

]block

inthe

topology.S

incethe

atomnum

bersin

everym

oleculetypein

thetopology

startat1and

thenum

bersin

theinputfile

forgenprnum

berconsecutively

from1,genpr

willonly

producea

usefulfilefor

thefirstm

olecule.F

iles-f

con

f.gro

InputG

enericstructure:

grog96

pdbtpr

tpbtpa

-nin

de

x.nd

xInput,O

pt.Index

file-o

po

sre.itp

Output

Includefile

fortopology

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int0

Setthe

nicelevel-fc

vector1

00

01

00

01

00

0force

constants(kJ

mol-1

nm-2)

E.53

gmxcheck

gmxcheck

readsa

trajectory(

.trj,.trr

or.xtc)oran

energyfile

(.en

eor.e

dr

)andprints

outusefulinform

ationaboutthem

.

Option

-cchecks

forpresence

ofcoordinates,velocitiesand

boxin

thefile,forclose

contacts(sm

allerthan-vd

wfa

cand

notbonded,

i.e.not

between-bo

nlo

and-b

on

hi

,all

relativeto

thesum

ofboth

Vander

Waals

radii)and

atoms

outsidethe

box(these

may

occuroften

andare

noproblem

).If

velocitiesare

present,anestim

atedtem

peraturew

illbecalculated

fromthem

.

The

programw

illcompare

runinput(

.tpr

,.tpb

or.tpa

)files

when

both-s1and

-s2are

supplied.

Page 33: CS USER AL Simulations - GROMACS

21

6A

pp

en

dix

E.

Ma

nu

alP

age

s

•M

olec

ules

mus

tbe

who

lein

the

initi

alco

nfigu

ratio

ns.

•A

tthe

mom

ent-

cion

lyw

orks

whe

nin

sert

ing

one

mol

ecul

e.

E.5

0ge

ncon

f

genc

onfm

ultip

lies

agi

ven

coor

dina

tefil

eby

sim

ply

stac

king

them

onto

pof

each

othe

r,lik

ea

smal

lchi

ldpl

ayin

gw

ithw

oode

nbl

ocks

.T

hepr

ogra

mm

akes

agr

idof

use

rd

efin

edp

ropo

rtio

ns(-n

bo

x),

and

inte

r-sp

aces

the

grid

poin

twith

anex

tra

spac

e-d

ist

.

Whe

nop

tion-

rot

isus

edth

epr

ogra

mdo

esno

tche

ckfo

rov

erla

pbe

twee

nm

olec

ules

ongr

idpo

ints

.It

isre

com

men

ded

tom

ake

the

box

inth

ein

putfi

leat

leas

tas

big

asth

eco

ordi

nate

s+

Van

der

Waa

lsra

dius

.

Ifth

eop

tiona

ltra

ject

ory

file

isgi

ven,

conf

orm

atio

nsar

eno

tgen

erat

ed,b

utre

adfr

omth

isfil

ean

dtr

ansl

ated

appr

opria

tely

tobu

ildth

egr

id.

File

s-f

con

f.g

roIn

put

Gen

eric

stru

ctur

e:gr

og9

6pd

btp

rtp

btp

a-o

ou

t.g

roO

utpu

tG

ener

icst

ruct

ure:

gro

g96

pdb

-trj

tra

j.xtc

Inpu

t,O

pt.

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-n

bo

xve

ctor

11

1N

umbe

rof

boxe

s-d

ist

vect

or0

00

Dis

tanc

ebe

twee

nbo

xes

-se

ed

int

0R

ando

mge

nera

tor

seed

,if0

gene

rate

dfr

omth

etim

e-r

ot

bool

no

Ran

dom

lyro

tate

conf

orm

atio

ns-s

hu

ffle

bool

no

Ran

dom

shuf

fling

ofm

olec

ules

-so

rtbo

oln

oS

ortm

olec

ules

onX

coor

d-b

lock

int

1D

ivid

eth

ebo

xin

bloc

kson

this

num

ber

ofcp

us-n

mo

lat

int

3N

umbe

rof

atom

spe

rm

olec

ule,

assu

med

tost

art

from

0.If

you

set

this

wro

ng,i

twill

scre

wup

your

syst

em!

-ma

xro

tve

ctor9

09

09

0M

axim

umra

ndom

rota

tion

-re

nu

mb

er

bool

no

Ren

umbe

rre

sidu

es

•T

hepr

ogra

msh

ould

allo

wfo

rra

ndom

disp

lace

men

toff

latti

cepo

ints

.

E.5

1ge

nion

geni

onre

plac

esso

lven

tmol

ecul

esby

mon

oato

mic

ions

atth

epo

sitio

nof

the

first

atom

sw

ithth

em

ostf

avor

-ab

leel

ectr

osta

ticpo

tent

ialo

rat

rand

om.

The

pote

ntia

lis

calc

ulat

edon

alla

tom

s,us

ing

norm

alG

RO

MA

CS

part

icle

base

dm

etho

ds(in

cont

rast

toot

her

met

hods

base

don

solv

ing

the

Poi

sson

-Bol

tzm

ann

equa

tion)

.T

hepo

tent

iali

sre

calc

ulat

edaf

ter

ever

yio

nin

sert

ion.

Ifsp

ecifi

edin

the

run

inpu

tfil

e,a

reac

tion

field

orsh

iftfu

nctio

nca

nbe

used

.T

hegr

oup

ofso

lven

tmol

ecul

essh

ould

beco

ntin

uous

and

allm

olec

ules

shou

ldha

veth

esa

me

num

ber

ofat

oms.

The

user

shou

ldad

dth

eio

nm

olec

ules

toth

eto

polo

gyfil

ean

din

clud

eth

efil

eio

ns.

itp.

Ion

nam

esfo

rG

rom

os96

shou

ldin

clud

eth

ech

arge

.

The

pote

ntia

lcan

bew

ritte

nas

B-f

acto

rsin

apd

bfil

e(f

orvi

sual

isat

ion

usin

ge.

g.ra

smol

).T

heun

itof

the

pote

ntia

lis

0.00

1kJ

/(m

ole)

.

For

larg

erio

ns,e

.g.

sulfa

tew

ere

com

men

ded

tous

ege

nbox

.

3.2

.P

erio

dic

bo

un

da

ryco

nd

itio

ns

13

Fig

ure

3.2:

Arh

ombi

cdo

deca

hedr

onan

dtr

unca

ted

octa

hedr

on(a

rbitr

ary

orie

ntat

ions

).

box

type

imag

ebo

xbo

xve

ctor

sbo

xve

ctor

angl

esdi

stan

cevo

lum

ea

bc

6bc

6ac

6ab

d0

0cu

bic

dd

30

d0

90◦

90◦

90◦

00

d

rhom

bic

d0

1 2d

dode

cahe

dron

d1 2

√2d

30

d1 2d

60◦

60◦

90◦

(xy-

squa

re)

≈0.

71d

30

01 2

√2d

rhom

bic

d1 2d

1 2d

dode

cahe

dron

d1 2

√2d

30

1 2

√3d

1 6

√3d

60◦

60◦

60◦

(xy-

hexa

gon)

≈0.

71d

30

01 3

√6d

trun

cate

dd

1 3d

-1 3d

octa

hedr

ond

4 9

√3d

30

2 3

√2d

1 3

√2d

≈71◦

≈71◦

≈71◦

≈0.

77d

30

01 3

√6d

Tabl

e3.

1:T

hecu

bic

box,

the

rhom

bic

dode

cahe

dron

and

the

trun

cate

doc

tahe

dron

.

the

outp

uttr

ajec

tory

itm

ight

seem

like

the

sim

ulat

ion

was

done

ina

rect

angu

larb

ox.

The

prog

ram

trjc

on

vca

nbe

used

toco

nver

tthe

traj

ecto

ryto

adi

ffere

ntun

it-ce

llre

pres

enta

tion.

Itis

also

poss

ible

tosi

mul

ate

with

out

perio

dic

boun

dary

cond

ition

s,bu

tit

ism

ore

effic

ient

tosi

mul

ate

anis

olat

edcl

uste

rof

mol

ecul

esin

ala

rge

perio

dic

box,

sinc

efa

stgr

idse

arch

ing

can

only

beus

edin

ape

riodi

csy

stem

.

3.2.

1S

ome

usef

ulbo

xty

pes

The

thre

em

ost

usef

ulbo

xty

pes

for

sim

ulat

ions

ofso

lvat

edsy

stem

sar

ede

scrib

edin

Tabl

e3.

1.T

herh

ombi

cdo

deca

hedr

on(F

ig.

3.2)

isth

esm

alle

stan

dm

ostr

egul

arsp

ace-

fillin

gun

itce

ll.E

ach

ofth

e12

imag

ece

llsis

atth

esa

me

dist

ance

.T

hevo

lum

eis

71%

ofth

evo

lum

eof

acu

bic

box

with

the

sam

eim

age

dist

ance

.T

his

save

sab

out

29%

ofC

PU

-tim

ew

hen

sim

ulat

ing

asp

heric

alor

flexi

ble

mol

ecul

ein

solv

ent.

Arh

ombi

cdo

deca

hedr

onca

nha

vetw

odi

ffere

ntor

ient

atio

ns,

whi

lefu

lfilli

ngeq

uatio

ns(3.

1).

The

prog

ram

ed

itco

nf

prod

uces

the

orie

ntat

ion

whi

chha

sa

Page 34: CS USER AL Simulations - GROMACS

14

Ch

ap

ter

3.

Alg

orith

ms

squarecross-section

with

thexy-plane.

This

orientationw

aschosen

becausethe

firsttw

obox

vectorscoincide

with

thex

andy-axis,

which

iseasier

tocom

prehend.T

heother

orientationcan

beusefulfor

simulations

ofmem

braneproteins.

Inthis

casethe

cross-sectionw

iththe

xy-planeis

ahexagon,w

hichhas

anarea

which

is14%

smaller

thanthe

areaofa

squarew

iththe

same

image

distance.T

heheight

ofthe

box(

cz )

shouldbe

changedto

obtainan

optimalspacing.

This

boxshape

doesnotonly

saveC

PU

-time,italso

resultsin

am

oreuniform

arrangementofthe

proteins.

3.2.2C

ut-offrestrictions

The

minim

umim

ageconvention

implies

thatthecut-offradius

usedto

truncatenon-bonded

inter-actions

mustnotexceed

halftheshortestbox

vectorfor

gridsearch:

Rc<

12m

in(‖a‖,‖b‖,‖c‖),

(3.4)

otherwise

more

thanone

image

would

bew

ithinthe

cut-offdistanceofthe

force.W

hena

macro-

molecule,

suchas

aprotein,

isstudied

insolution,

thisrestriction

doesnot

suffice.In

principlea

singlesolventm

oleculeshould

notbeable

to‘see’both

sidesofthe

macrom

olecule.T

hism

eansthatthe

lengthofeach

boxvector

mustexceed

thelength

ofthem

acromolecule

inthe

directionof

thatedgep

lus

two

times

thecut-off

radiusRc .

Itis

comm

onto

comprom

isein

thisrespect,

andm

akethe

solventlayersom

ewhatsm

allerin

orderto

reducethe

computationalcost.

For

efficiencyreasons

thecut-offw

ithsim

plesearch

intriclinic

boxes(grid

searchalw

aysuses

eq.(

3.4))is

more

restriced:

Rc<

12m

in(ax ,b

y ,cz )

(3.5)

Each

unitcell

(cubic,rectangular

ortriclinic)

issurrounded

by26

translatedim

ages.T

husa

particularim

agecan

always

beidentified

byan

indexpointing

toone

of27tra

nsla

tion

vecto

rsandconstructed

byapplying

atranslation

with

theindexed

vector(see3.4.3).

3.3T

hegroup

concept

Inthe

GR

OM

AC

SM

Dand

analysisprogram

sone

usesg

rou

psofatom

sto

performcertain

actionson.

The

maxim

umnum

berof

groupsis

256,but

everyatom

canonly

belongto

fourdifferent

groups,oneofeach

ofthefollow

ingkinds:

T-couplinggroup

The

temperature

couplingparam

eters(reference

temperature,

time

constant,num

berof

degreesof

freedom,

see3.4.4)

canbe

definedfor

eachT-coupling

groupsepa-

rately.F

orexam

ple,in

asolvated

macrom

oleculethe

solvent(that

tendsto

producem

oreheating

byforce

andintegration

errors)canbe

coupledw

itha

shortertime

constanttoa

baththan

am

acromolecule,

ora

surfacecan

bekept

coolerthan

anadsorbing

molecule.

Many

differentT-couplinggroups

may

bedefined.

Freeze

groupA

toms

thatbelong

toa

freezegroup

arekept

stationaryin

thedynam

ics.T

hisis

usefulduringequilibration,

e.g.to

avoidthat

badlyplaced

solventm

oleculesw

illgiveun-

reasonablekicks

toprotein

atoms,although

thesam

eeffectcan

alsobe

obtainedby

puttinga

restrainingpotentialon

theatom

sthat

must

beprotected.

The

freezeoption

canbe

used

E.4

9.

gen

bo

x2

15

E.49

genbox

Genbox

cando

oneof3

things:

1)G

eneratea

boxofsolvent.

Specify

-csand

-box.O

rspecify

-csand

-cpw

itha

structurefile

with

abox,

butwithoutatom

s.

2)S

olvatea

soluteconfiguration,

eg.a

protein,in

abath

ofsolvent

molecules.

Specify-cp

(solute)and

-cs(solvent).

The

boxspecified

inthe

solutecoordinate

file(

-cp)

isused,

unless-bo

xis

set,w

hichalso

centersthe

solute.T

heprogramed

itcon

fhas

more

sophisticatedoptions

tochange

thebox

andcenter

thesolute.

Solventm

oleculesare

removed

fromthe

boxw

herethe

distancebetw

eenany

atomofthe

solutem

olecule(s)and

anyatom

ofthe

solventm

oleculeis

lessthan

thesum

ofthe

VanderWaals

radiiofboth

atoms.

Adatabase

(vd

wra

dii.d

at

)ofVanderW

aalsradiiis

readby

theprogram

,atoms

notinthe

databaseare

assigneda

defaultdistance-vd

w.

3)Insertanum

ber(-nmo

l)ofextra

molecules

(-ci)atrandom

positions.T

heprogram

iteratesuntiln

mo

lm

oleculeshave

beeninserted

inthe

box.To

testwhether

aninsertion

issuccessfulthe

same

VanderWaals

criteriumis

usedas

forrem

ovalofsolvent

molecules.

When

noappropriately

sizedholes

(holesthat

canhold

anextra

molecule)

areavailable

theprogram

triesfor

-nm

ol

*-try

times

beforegiving

up.Increase

-tryifyou

haveseveralsm

allholesto

fill.

The

defaultsolventisS

imple

PointC

hargew

ater(S

PC

),with

coordinatesfrom

$G

MX

LIB

/spc2

16

.gro

.O

thersolvents

arealso

supported,asw

ellasm

ixedsolvents.

The

onlyrestriction

tosolventtypes

isthata

solventmolecule

consistsofexactly

oneresidue.

The

residueinform

ationin

thecoordinate

filesis

used,andshould

thereforebe

more

orless

consistent.In

practicethis

means

thattw

osubsequent

solventm

oleculesin

thesolvent

coordinatefile

shouldhave

differentresidue

number.

The

boxof

soluteis

builtby

stackingthe

coordinatesread

fromthe

coordinatefile.

This

means

thatthese

coordinatesshould

beequlibrated

inperiodic

boundaryconditions

toensure

agood

alignmentofm

oleculeson

thestacking

interfaces.

The

programcan

optionallyrotate

thesolute

molecule

toalign

thelongestm

oleculeaxis

alonga

boxedge.

This

way

theam

ountof

solventm

oleculesnecessary

isreduced.

Itshould

bekept

inm

indthat

thisonly

works

forshort

simulations,

aseg.

analpha-helicalpeptide

insolution

canrotate

over90

degrees,w

ithin500

ps.In

generalitistherefore

betterto

make

am

oreor

lesscubic

box.

Setting

-shelllargerthan

zerow

illplacea

layerofw

aterofthe

specifiedthickness

(nm)

aroundthe

solute.H

int:itis

agood

ideato

puttheprotein

inthe

centerofa

boxfirst(using

editconf).

Finally,genbox

willoptionally

remove

linesfrom

yourtopologyfile

inw

hicha

numberofsolventm

oleculesis

alreadyadded,and

addsa

linew

iththe

totalnumber

ofsolventmolecules

inyour

coordinatefile.

Files-cp

pro

tein

.gro

Input,Opt.

Generic

structure:gro

g96pdb

tprtpb

tpa-cs

spc2

16

.gro

Input,Opt.,Lib.Generic

structure:gro

g96pdb

tprtpb

tpa-ci

inse

rt.gro

Input,Opt.

Generic

structure:gro

g96pdb

tprtpb

tpa-o

ou

t.gro

Output

Generic

structure:gro

g96pdb

-pto

po

l.top

In/Out,O

pt.Topology

file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-bo

xvector

00

0box

size-n

mo

lint

0no

ofextram

oleculesto

insert-try

int1

0try

inserting-nm

ol*-trytim

es-se

ed

int1

99

7random

generatorseed

-vdw

dreal

0.1

05

defaultvdwaals

distance-sh

ell

real0

thicknessofoptionalw

aterlayer

aroundsolute

Page 35: CS USER AL Simulations - GROMACS

21

4A

pp

en

dix

E.

Ma

nu

alP

age

s

-ot

tem

p.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-e

kre

kro

t.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-tu

enum

ps

Tim

eun

it:p

s,f

s,n

s,u

s,m

s,s

,mor

h-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-co

mbo

oln

oP

lotd

ata

for

the

com

ofea

chgr

oup

-mo

lbo

oln

oIn

dex

cont

ains

mol

ecul

enu

mbe

rsis

oat

omnu

mbe

rs-n

oju

mp

bool

no

Rem

ove

jum

psof

atom

sac

ross

the

box

-xbo

olye

sP

lotX

-com

pone

nt-y

bool

yes

Plo

tY-c

ompo

nent

-zbo

olye

sP

lotZ

-com

pone

nt-le

nbo

oln

oP

lotv

ecto

rle

ngth

E.4

8g

vela

cc

gve

lacc

com

pute

sth

eve

loci

tyau

toco

rrel

atio

nfu

nctio

n.W

hen

the

-sop

tion

isus

ed,

the

mom

entu

mau

toco

rrel

atio

nfu

nctio

nis

calc

ulat

ed.

With

optio

n-m

ol

the

mom

entu

mau

toco

rrel

atio

nfu

nctio

nof

mol

ecul

esis

calc

ulat

ed.

Inth

isca

seth

ein

dex

grou

psh

ould

cons

isto

fmol

ecul

enu

mbe

rsin

stea

dof

atom

num

bers

.

File

s-f

tra

j.trr

Inpu

tF

ullp

reci

sion

traj

ecto

ry:

trr

trj

-sto

po

l.tp

rIn

put,

Opt

.S

truc

ture

+m

ass(

db):

tpr

tpb

tpa

gro

g96

pdb

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e-o

vac.

xvg

Out

put

xvgr

/xm

grfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-m

ol

bool

no

Cal

cula

teva

cof

mol

ecul

es-a

cfle

nin

t-1

Leng

thof

the

AC

F,de

faul

tis

half

the

num

ber

offr

ames

-no

rma

lize

bool

yes

Nor

mal

ize

AC

F-P

enum

0O

rder

ofLe

gend

repo

lyno

mia

lfor

AC

F(0

indi

cate

sno

ne):

0,1

,2or

3-f

itfn

enum

no

ne

Fit

func

tion:

no

ne

,exp

,ae

xp,e

xpe

xpor

vac

-ncs

kip

int

0S

kip

Npo

ints

inth

eou

tput

file

ofco

rrel

atio

nfu

nctio

ns-b

eg

infit

real

0T

ime

whe

reto

begi

nth

eex

pone

ntia

lfito

fthe

corr

elat

ion

func

tion

-en

dfit

real

-1T

ime

whe

reto

end

the

expo

nent

ialfi

toft

heco

rrel

atio

nfu

nctio

n,-1

istil

lth

een

d

3.4

.M

ole

cula

rD

yna

mic

s1

5

onon

eor

two

coor

dina

tes

ofan

atom

,th

ereb

yfr

eezi

ngth

eat

oms

ina

plan

eor

ona

line.

Man

yfr

eeze

grou

psca

nbe

defin

ed.

Acc

eler

ate

grou

pO

nea

chat

omin

an’a

ccel

erat

egr

oup’

anac

cele

ratio

na

gw

illbe

impo

sed.

Thi

sis

equi

vale

ntto

anex

tern

alfo

rce.

Thi

sfe

atur

em

akes

itpo

ssib

leto

driv

eth

esy

stem

into

ano

n-eq

uilib

rium

stat

ean

den

able

sto

perf

orm

non-

equi

libriu

mM

Dto

obta

intr

ansp

ort

prop

ertie

s.

Ene

rgy

mon

itor

grou

pM

utua

lint

erac

tions

betw

een

alle

nerg

ym

onito

rgro

ups

are

com

pile

ddu

r-in

gth

esi

mul

atio

n.T

his

isdo

nese

para

tely

for

Lenn

ard-

Jone

san

dC

oulo

mb

term

s.In

prin

-ci

ple

upto

256

grou

psco

uld

bede

fined

,but

that

wou

ldle

adto

256

×25

6ite

ms!

Bet

ter

use

this

conc

epts

parin

gly.

All

non-

bond

edin

tera

ctio

nsbe

twee

npa

irsof

ener

gym

onito

rgr

oups

can

beex

clud

ed(s

eese

c.7.

3.1)

.P

airs

ofpa

rtic

les

from

excl

uded

pairs

ofen

ergy

mon

itor

grou

psar

eno

tpu

tin

toth

epa

irlis

t.T

his

can

resu

ltin

asi

gnifi

cant

spee

dup

for

sim

ulat

ions

whe

rein

tera

ctio

nsw

ithin

orbe

twee

npa

rts

ofth

esy

stem

are

notr

equi

red.

The

use

ofgr

oups

inan

alys

ispr

ogra

ms

isde

scrib

edin

chap

ter

8.

3.4

Mol

ecul

arD

ynam

ics

Agl

obal

flow

sche

me

for

MD

isgi

ven

inF

ig.3.3.

Eac

hM

Dor

EM

run

requ

ires

asin

puta

seto

fin

itial

coor

dina

tes

and

-op

tiona

lly-

initi

alve

loci

ties

ofal

lpar

ticle

sin

volv

ed.

Thi

sch

apte

rdo

esno

tdes

crib

eho

wth

ese

are

obta

ined

;for

the

setu

pof

anac

tual

MD

run

chec

kth

eon

line

man

uala

tw

ww

.gro

mac

s.or

g.

3.4.

1In

itial

cond

ition

s

Topo

logy

and

forc

efie

ld

The

syst

emto

polo

gy,

incl

udin

ga

desc

riptio

nof

the

forc

efie

ld,

mus

tbe

load

ed.

The

seite

ms

are

desc

ribed

inch

apte

r4.A

llth

isin

form

atio

nis

stat

ic;i

tis

neve

rm

odifi

eddu

ring

the

run.

Coo

rdin

ates

and

velo

citie

s

The

n,be

fore

aru

nst

arts

,th

ebo

xsi

zean

dth

eco

ordi

nate

san

dve

loci

ties

ofal

lpar

ticle

sar

ere

-qu

ired.

The

box

size

isde

term

ined

byth

ree

vect

ors

(nin

enu

mbe

rs)

b1,b

2,b

3,w

hich

repr

esen

tthe

thre

eba

sis

vect

ors

ofth

epe

riodi

cbo

x.W

hile

inth

epr

esen

tve

rsio

nof

GR

OM

AC

Son

lyre

ctan

-gu

lar

boxe

sar

eal

low

ed,

thre

enu

mbe

rssu

ffice

,bu

tth

eus

eof

thre

eve

ctor

sal

read

ypr

epar

esfo

rar

bitr

ary

tric

linic

boxe

sto

beim

plem

ente

din

ala

ter

vers

ion.

Ifth

eru

nst

arts

att=t 0

,th

eco

ordi

nate

satt

=t 0

mus

tbe

know

n.T

hele

ap

-fro

ga

lgo

rith

m,

used

toup

date

the

time

step

with∆t

(see

3.4.

4),

requ

ires

that

the

velo

citie

sm

ust

bekn

own

att

=t 0−

∆t

2.

Ifve

loci

ties

are

not

avai

labl

e,th

epr

ogra

mca

nge

nera

tein

itial

atom

icve

loci

ties

Page 36: CS USER AL Simulations - GROMACS

16

Ch

ap

ter

3.

Alg

orith

ms

TH

EG

LOB

AL

MD

ALG

OR

ITH

M

1.Inputinitialconditions

PotentialinteractionV

asa

functionofatom

positionsP

ositionsrofallatom

sin

thesystem

Velocitiesv

ofallatoms

inthe

system⇓

repeat2,3,4requirednum

berofsteps:

2.C

ompute

forces

The

forceon

anyatom

Fi =

−∂V

∂ri

iscom

putedby

calculatingthe

forcebetw

eennon-bonded

atompairs:

Fi = ∑

jF

ij

plusthe

forcesdue

tobonded

interactions(w

hichm

aydepend

on1,

2,3,or4

atoms),plus

restrainingand/or

externalforces.T

hepotentialand

kineticenergies

andthe

pressuretensor

arecom

puted.⇓

3.U

pdateconfiguration

The

movem

entoftheatom

sis

simulated

bynum

ericallysolving

New

ton’sequations

ofmotion

d2r

i

dt 2=

Fi

mi

ordr

i

dt=

vi ;

dv

i

dt

=F

i

mi

⇓4.ifrequired:O

utputstepw

ritepositions,velocities,energies,tem

perature,pressure,etc.

Figure

3.3:T

heglobalM

Dalgorithm

E.4

7.

gtra

j2

13

When

thebox

iscubic,

onecan

usethe

option-o

c,

which

averagesthe

tcaf’sover

allk-vectorsw

iththe

same

length.T

hisresults

inm

oreaccurate

tcaf’s.B

oththe

cubictcaf’s

andfits

arew

rittento

-oc

The

cubiceta

estimates

arealso

written

to-o

v.

With

option-m

ol

thetransverse

currentisdeterm

inedofm

oleculesinstead

ofatoms.

Inthis

casethe

indexgroup

shouldconsistofm

oleculenum

bersinstead

ofatomnum

bers.

The

k-dependentviscositiesin

the-ovfile

shouldbe

fittedto

eta(k)=

eta0(1

-a

k2)

toobtain

theviscosity

atinfinitew

avelength.

NO

TE

:make

sureyou

write

coordinatesand

velocitiesoften

enough.T

heinitial,

non-exponential,partof

theautocorrelation

functionis

veryim

portantforobtaining

agood

fit.

Files

-ftra

j.trrInput

Fullprecision

trajectory:trr

trj-s

top

ol.tp

rInput,O

pt.S

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-n

ind

ex.n

dx

Input,Opt.

Indexfile

-ot

tran

scur.xvg

Output,O

pt.xvgr/xm

grfile

-oa

tcaf

all.xvg

Output

xvgr/xmgr

file-o

tcaf.xvg

Output

xvgr/xmgr

file-o

ftca

ffit.xvg

Output

xvgr/xmgr

file-o

ctca

fcu

b.xvg

Output,O

pt.xvgr/xm

grfile

-ov

visck.xvg

Output

xvgr/xmgr

file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-m

ol

booln

oC

alculatetcafofm

olecules-k3

4bool

no

Also

usek=

(3,0,0)and

k=(4,0,0)

-wt

real5

Exponentialdecay

time

forthe

TC

AF

fitweights

E.47

gtraj

gtrajplots

coordinates,velocities,forcesand/or

thebox.

With

-com

thecoordinates,velocities

andforces

arecalculated

forthe

centerof

mass

ofeach

group.W

hen-m

ol

isset,

thenum

bersin

theindex

fileare

interpretedas

molecule

numbers

andthe

same

procedureas

with

-com

isused

foreach

molecule.

Option

-ot

plotsthe

temperature

ofeach

group,provided

velocitiesare

presentin

thetrajectory

file.This

implies

-com

.

Option

-ekr

plotsthe

rotationalkineticenergy

ofeachgroup,provided

velocitiesare

presentinthe

trajec-tory

file.T

hisim

plies-com

.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-n

ind

ex.n

dx

Input,Opt.

Indexfile

-ox

coo

rd.xvg

Output,O

pt.xvgr/xm

grfile

-ov

velo

c.xvgO

utput,Opt.

xvgr/xmgr

file-o

ffo

rce.xvg

Output,O

pt.xvgr/xm

grfile

-ob

bo

x.xvgO

utput,Opt.

xvgr/xmgr

file

Page 37: CS USER AL Simulations - GROMACS

21

2A

pp

en

dix

E.

Ma

nu

alP

age

s

E.4

5g

sorie

nt

gso

rient

anal

yzes

solv

ento

rient

atio

nar

ound

solu

tes.

Itca

lcul

ates

two

angl

esbe

twee

nth

eve

ctor

from

one

orm

ore

refe

renc

epo

sitio

nsto

the

first

atom

ofea

chso

lven

tmol

ecul

e:th

eta1

:th

ean

gle

with

the

vect

orfr

omth

efir

stat

omof

the

solv

entm

olec

ule

toth

em

idpo

intb

etw

een

atom

s2

and

3.th

eta2

:th

ean

gle

with

the

norm

alof

the

solv

entp

lane

,defi

ned

byth

esa

me

thre

eat

oms.

The

refe

renc

eca

nbe

ase

tof

atom

sor

the

cent

erof

mas

sof

ase

tof

atom

s.T

hegr

oup

ofso

lven

tat

oms

shou

ldco

nsis

tof

3at

oms

per

solv

ent

mol

ecul

e.O

nly

solv

ent

mol

ecul

esbe

twee

n-r

min

and

-rm

ax

are

cons

ider

edea

chfr

ame.

-o:

angl

edi

strib

utio

nof

thet

a1.

-no

:an

gle

dist

ribut

ion

ofth

eta2

.

-ro

:<

cos(

thet

a1)>

and<

3cos

2(t

heta

2)-1>

asa

func

tion

ofth

edi

stan

ce.

-ro

:th

esu

mov

eral

lsol

vent

mol

ecul

esw

ithin

dist

ance

rof

cos(

thet

a1)

and

3cos

2(t

heta

2)-1

asa

func

tion

ofr.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Str

uctu

re+

mas

s(db

):tp

rtp

btp

agr

og9

6pd

b-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-oso

ri.x

vgO

utpu

txv

gr/x

mgr

file

-no

sno

r.xv

gO

utpu

txv

gr/x

mgr

file

-ro

sord

.xvg

Out

put

xvgr

/xm

grfil

e-c

osc

um

.xvg

Out

put

xvgr

/xm

grfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-c

om

bool

no

Use

the

cent

erof

mas

sas

the

refe

renc

epo

stio

n-r

min

real

0M

inim

umdi

stan

ce-r

ma

xre

al0

.5M

axim

umdi

stan

ce-n

bin

int

20

Num

ber

ofbi

ns

E.4

6g

tcaf

gtc

afco

mpu

tes

tran

vers

ecu

rren

taut

ocor

rela

tions

.T

hese

are

used

toes

timat

eth

esh

ear

visc

osity

eta.

For

deta

ilsse

e:P

alm

er,J

CP

49(1

994)

pp35

9-36

6.

Tra

nsve

rse

curr

ents

are

calc

ulat

edus

ing

the

k-ve

ctor

s(1

,0,0

)an

d(2

,0,0

)ea

chal

soin

the

y-an

dz-

dire

ctio

n,(1

,1,0

)an

d(1

,-1,

0)ea

chal

soin

the

2ot

her

plai

ns(t

hese

vect

ors

are

noti

ndep

ende

nt)

and

(1,1

,1)

and

the

3ot

her

box

diag

onal

s(a

lso

noti

ndep

ende

nt).

For

each

k-ve

ctor

the

sine

and

cosi

near

eus

ed,i

nco

mbi

natio

nw

ithth

eve

loci

tyin

2pe

rpen

dicu

lar

dire

ctio

ns.

Thi

sgi

ves

ato

talo

f16

*2*2

=64

tran

sver

secu

rren

ts.

One

auto

corr

elat

ion

isca

lcul

ated

fitte

dfo

rea

chk-

vect

or,w

hich

give

s16

tcaf

’s.

Eac

hof

thes

etc

af’s

isfit

ted

tof(

t)=

exp(

-v)(

cosh

(Wv)

+1/

Wsi

nh(W

v)),

v=

-t/(

2ta

u),W

=sq

rt(1

-4

tau

eta/

rho

k2),

whi

chgi

ves

16ta

u’s

and

eta’

s.T

hefit

wei

ghts

deca

yw

ithtim

eas

exp(

-t/w

t),t

hetc

afan

dfit

are

calc

ulat

edup

totim

e5*

wt.

The

eta’

ssh

ould

befit

ted

to1

-a

eta(

k)k

2,f

rom

whi

chon

eca

nes

timat

eth

esh

ear

visc

osity

atk=

0.

3.4

.M

ole

cula

rD

yna

mic

s1

7

0�

Vel

ocity

0.0

Probability �

Fig

ure

3.4:

AM

axw

ellia

ndi

strib

utio

n,ge

nera

ted

from

rand

omnu

mbe

rs.

v i,i

=1...3N

from

aM

axw

ellia

ndi

strib

utio

n(F

ig.3.

4)at

agi

ven

abso

lute

tem

pera

tureT:

p(v

i)=√ m

i

2πkT

exp(−m

iv2 i

2kT

)(3

.6)

whe

rek

isB

oltz

man

n’s

cons

tant

(see

chap

ter

2).

Toac

com

plis

hth

is,n

orm

ally

dist

ribut

edra

ndom

num

bers

are

gene

rate

dby

addi

ngtw

elve

rand

omnu

mbe

rsR

kin

the

rang

e0≤R

k<

1an

dsu

btra

ctin

g6.

0fr

omth

eir

sum

.T

here

sult

isth

enm

ultip

lied

byth

est

anda

rdde

viat

ion

ofth

eve

loci

tydi

strib

utio

n√ kT/m

i.S

ince

the

resu

lting

tota

lene

rgy

will

notc

orre

spon

dex

actly

toth

ere

quire

dte

mpe

ratu

reT,a

corr

ectio

nis

mad

e:fir

stth

ece

nter

-of-

mas

sm

otio

nis

rem

oved

and

then

allv

eloc

ities

are

scal

edsu

chth

atth

eto

tale

nerg

yco

rres

pond

sex

actly

toT

(see

eqn.3

.12)

.

Cen

ter-

of-m

ass

mot

ion

The

cent

er-o

f-m

ass

velo

city

isno

rmal

lyse

tto

zero

atev

ery

step

.N

orm

ally

ther

eis

none

text

erna

lfo

rce

actin

gon

the

syst

eman

dth

ece

nter

-of-

mas

sve

loci

tysh

ould

rem

ain

cons

tant

.In

prac

tice,

how

ever

,th

eup

date

algo

rithm

deve

lops

ave

rysl

owch

ange

inth

ece

nter

-of-

mas

sve

loci

ty,

and

thus

inth

eto

talk

inet

icen

ergy

ofth

esy

stem

,spe

cial

lyw

hen

tem

pera

ture

coup

ling

isus

ed.

Ifsu

chch

ange

sar

eno

tque

nche

d,an

appr

ecia

ble

cent

er-o

f-m

ass

mot

ion

deve

lops

even

tual

lyin

long

runs

,an

dth

ete

mpe

ratu

rew

illbe

sign

ifica

ntly

mis

inte

rpre

ted.

The

sam

em

ayha

ppen

due

toov

eral

lro

tatio

nalm

otio

n,bu

ton

lyw

hen

anis

olat

edcl

uste

ris

sim

ulat

ed.

Inpe

riodi

csy

stem

sw

ithfil

led

boxe

s,th

eov

eral

lrot

atio

nalm

otio

nis

coup

led

toot

her

degr

ees

offr

eedo

man

ddo

esno

tgiv

ean

ypr

oble

ms.

Page 38: CS USER AL Simulations - GROMACS

18

Ch

ap

ter

3.

Alg

orith

ms

3.4.2N

eighborsearching

As

mentioned

inchapter4,

internalforcesare

eithergenerated

fromfixed

(static)lists,

orfrom

dynamics

lists.T

helatter

concernnon-bonded

interactionsbetw

eenany

pairof

particles.W

hencalculating

thenon-bonded

forces,it

isconvenient

tohave

allparticlesin

arectangular

box.A

sshow

nin

Fig.3.1,

itis

possibleto

transforma

triclinicbox

intoa

rectangularbox.

The

outputcoordinates

arealw

aysin

arectangular

box,evenw

hena

dodecahedronor

triclinicbox

was

usedfor

thesim

ulation.E

quations(

3.1)ensure

thatwe

canresetparticles

ina

rectangularbox

byfirst

shiftingthem

with

boxvectorc,then

withb

andfinally

witha

.E

quations(3.3)ensure

thatwe

canfind

the14

nearesttriclinic

images

within

alinear

combination

which

doesnot

involvem

ultiplesofbox

vectors.

Pair

listsgeneration

The

non-bondedpair

forcesneed

tobe

calculatedonly

forthose

pairsi,j

forw

hichthe

distancerij

betweeni

andthe

nearestimage

ofj

isless

thana

givencut-offradiusRc .

Som

eofthe

particlepairs

thatfulfillthiscriterion

areexcluded,w

hentheir

interactionis

alreadyfully

accountedfor

bybonded

interactions.G

RO

MA

CS

employs

ap

air

listthat

containsthose

particlepairs

forw

hichnon-bonded

forcesm

ustbecalculated.

The

pairlistcontains

theparticle

numbers

andan

indexfor

theim

agedisplacem

entvectors

thatm

ustbe

appliedto

obtainthe

nearestim

age,for

allparticlepairs

thathave

anearest-im

agedistance

lessthanrsh

ort

.T

helist

isupdated

everynstliststeps,

wheren

stlistis

typically10

forthe

GR

OM

AC

Sforcefield

and5

forthe

GR

OM

OS

-96forcefield.

There

isan

optionto

calculatethe

totalnon-bondedforce

oneach

particledue

toall

particlein

ashellaround

thelist-cutoff,

i.e,atadistance

betweenrsh

ort

andrlo

ng

.T

hisforce

iscalculated

duringthe

pairlistupdate

andretained

duringn

stliststeps.

Tom

akethe

neighborlistallparticlesthatare

close(i.e.w

ithinthe

cut-off)toa

givenparticle

must

befound.

This

searching,usually

calledneighbor

searching(N

S),

involvesperiodic

boundaryconditions

anddeterm

iningtheim

age(see

sec.3.2).W

ithoutperiodic

boundaryconditions

asim

pleO

(N2)

algorithmm

ustbe

used.W

ithperiodic

boundaryconditions

agrid

searchcan

beused,w

hichisO

(N).

Sim

plesearch

Due

toequations

(3.1)and

(3.5),the

vectorrij

connectingim

agesw

ithinthe

cut-offRccan

befound

byconstructing:

r′′′

=r

j −r

i(3.7)

r′′

=r′′′−

a∗

rou

nd

(r ′′′z/c

z ))(3.8)

r′

=r′′−

b∗

rou

nd

(r ′′y /by )

(3.9)

rij

=r′−

c∗

rou

nd

(r ′x /a

x )(3.10)

When

distancesbetw

eenany

two

particlesin

atriclinic

boxare

needed,m

anyshifts

ofcombina-

tionsofbox

vectorsneed

tobe

consideredto

findthe

nearestimage.

E.4

4.

gsg

an

gle

21

1

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-so

lsizereal

0.1

4R

adiusofthe

solventprobe(nm

)-n

do

tsint

24

Num

berofdots

persphere,m

oredots

means

more

accuracy-q

ma

xreal

0.2

The

maxim

umcharge

(e,absolutevalue)

ofahydrophobic

atom-m

ina

rea

real0

.5T

hem

aximum

charge(e,absolute

value)ofa

hydrophobicatom

-skipint

1D

oonly

everynth

frame

-pro

tbool

yes

Outputthe

proteinto

theconnelly

pdbfile

too

E.44

gsgangle

Com

putethe

angleand

distancebetw

eentw

ogroups.

The

groupsare

definedby

anum

berofatom

sgiven

inan

indexfile

andm

aybe

two

orthree

atoms

insize.

The

anglescalculated

dependon

theorder

inw

hichthe

atoms

aregiven.

Giving

forinstance

56

willrotate

thevector

5-6w

ith180

degreescom

paredto

giving6

5.

Ifthreeatom

sare

given,thenorm

alonthe

planespanned

bythose

threeatom

sw

illbecalculated,using

theform

ulaP

1P2

xP

1P3.

The

cosofthe

angleis

calculated,usingthe

inproductofthetw

onorm

alizedvectors.

Here

isw

hatsome

ofthefile

optionsdo:

-oa:A

nglebetw

eenthe

two

groupsspecified

inthe

indexfile.

Ifa

groupcontains

threeatom

sthe

normal

tothe

planedefined

bythose

threeatom

sw

illbeused.

Ifagroup

containstw

oatom

s,thevector

definedby

thosetw

oatom

sw

illbeused.

-od:D

istancebetw

eentw

ogroups.

Distance

istaken

fromthe

centerofone

groupto

thecenter

oftheother

group.-od1:

Ifoneplane

andone

vectoris

given,thedistances

foreach

oftheatom

sfrom

thecenter

oftheplane

isgiven

seperately.-od2:

For

two

planesthis

optionhas

nom

eaning.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-nin

de

x.nd

xInput

Indexfile

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-oa

sga

ng

le.xvg

Output

xvgr/xmgr

file-o

dsg

dist.xvg

Output

xvgr/xmgr

file-o

d1

sgd

ist1.xvg

Output

xvgr/xmgr

file-o

d2

sgd

ist2.xvg

Output

xvgr/xmgr

file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files

Page 39: CS USER AL Simulations - GROMACS

21

0A

pp

en

dix

E.

Ma

nu

alP

age

s

-acf

len

int

-1Le

ngth

ofth

eA

CF,

defa

ulti

sha

lfth

enu

mbe

rof

fram

es-n

orm

aliz

ebo

olye

sN

orm

aliz

eA

CF

-Pen

um0

Ord

erof

Lege

ndre

poly

nom

ialf

orA

CF

(0in

dica

tes

none

):0

,1,2

or3

-fitf

nen

umn

on

eF

itfu

nctio

n:n

on

e,e

xp,a

exp

,exp

exp

orva

c-n

cski

pin

t0

Ski

pN

poin

tsin

the

outp

utfil

eof

corr

elat

ion

func

tions

-be

gin

fitre

al0

Tim

ew

here

tobe

gin

the

expo

nent

ialfi

toft

heco

rrel

atio

nfu

nctio

n-e

nd

fitre

al-1

Tim

ew

here

toen

dth

eex

pone

ntia

lfito

fthe

corr

elat

ion

func

tion,

-1is

till

the

end

E.4

2g

saltb

r

gsa

ltbr

plot

sth

edi

ffere

nce

betw

een

allc

ombi

natio

nof

char

ged

grou

psas

afu

nctio

nof

time.

The

grou

psar

eco

mbi

ned

indi

ffere

ntw

ays.

Am

inim

umdi

stan

ceca

nbe

give

n,(e

g.th

ecu

t-of

f),

then

grou

psth

atar

ene

ver

clos

erth

anth

atdi

stan

cew

illno

tbe

plot

ted.

Out

put

will

bein

anu

mbe

rof

fixed

filen

ames

,m

in-m

in.x

vg,m

in-p

lus.

xvg

and

plus

-plu

s.xv

g,or

files

for

ever

yin

divi

dual

ion-

pair

ifse

lect

ed

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-tre

al1

00

0tr

unc

dist

ance

-se

pbo

oln

oU

sese

para

tefil

esfo

rea

chin

tera

ctio

n(m

aybe

MA

NY

)

E.4

3g

sas

gsa

sco

mpu

tes

hydr

opho

bic

and

tota

lsol

vent

acce

ssib

lesu

rfac

ear

ea.

As

asi

deef

fect

the

Con

nolly

surf

ace

can

bege

nera

ted

asw

elli

na

pdb

file

whe

reth

eno

des

are

repr

esen

ted

asat

oms

and

the

vert

ices

conn

ectin

gth

ene

ares

tno

des

asC

ON

EC

Tre

cord

s.T

hear

eaca

nbe

plot

ted

per

atom

and

per

resi

due

asw

ell(

optio

n-a

o).

Inco

mbi

natio

nw

ithth

ela

tter

optio

nanit

pfil

eca

nbe

gene

rate

d(o

ptio

n-i)

whi

chca

nbe

used

tore

stra

insu

rfac

eat

oms.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a-o

are

a.x

vgO

utpu

txv

gr/x

mgr

file

-rre

sare

a.x

vgO

utpu

txv

gr/x

mgr

file

-qco

nn

elly

.pd

bO

utpu

t,O

pt.

Pro

tein

data

bank

file

-ao

ato

ma

rea

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e-i

surf

at.itp

Out

put,

Opt

.In

clud

efil

efo

rto

polo

gy

3.4

.M

ole

cula

rD

yna

mic

s1

9

�����

�����

����

����

���������

���������

���������

��������

��������

��������

����������

����������

����������

����������

����������

����������

����������

����������

����������

����������

�������

�������

������

������ �

��

���

j

i

i’

���������������

�����

�����

�����

�����

�����

�����

�����

�����

Fig

ure

3.5:

Grid

sear

chin

two

dim

ensi

ons.

The

arro

ws

are

the

box

vect

ors.

Grid

sear

ch

The

grid

sear

chis

sche

mat

ical

lyde

pict

edin

Fig

.3.

5.A

llpa

rtic

les

are

puto

nth

eN

Sgr

id,w

ithth

esm

alle

stsp

acin

g≥R

c/2

inea

chof

the

dire

ctio

ns.

Inth

edi

rect

ion

ofea

chbo

xve

ctor

,a

part

icle

iha

sth

ree

imag

es.

For

each

dire

ctio

nth

eim

age

may

be-1

,0or

1,co

rres

pond

ing

toa

tran

slat

ion

over

-1,

0or

+1

box

vect

or.

We

dono

tse

arch

the

surr

ound

ing

NS

grid

cells

for

neig

hbor

sof

ian

dth

enca

lcul

ate

the

imag

e,bu

tra

ther

cons

truc

tth

eim

ages

first

and

then

sear

chne

ighb

ors

corr

espo

ndin

gto

that

imag

eofi.

As

can

been

seen

inF

ig.

3.5,

for

som

eim

ages

ofith

esa

me

grid

cell

mig

htbe

sear

ched

.T

his

isno

ta

prob

lem

,si

nce

atm

ost

one

imag

ew

ill“s

ee”

the

j-pa

rtic

le,

due

toth

em

inim

umim

age

conv

entio

n.F

orev

ery

part

icle

,le

ssth

an12

5(5

3)

neig

hbor

ing

cells

are

sear

ched

.T

here

fore

,th

eal

gorit

hmsc

ales

linea

rw

ithth

enu

mbe

rof

part

icle

s.A

lthou

ghth

epr

efac

tor

isla

rge

the

scal

ing

beha

vior

mak

esth

eal

gorit

hmfa

rsu

perio

rov

erth

est

anda

rdO

(N2)

algo

rithm

whe

nth

enu

mbe

rof

part

icle

sex

ceed

sa

few

hund

red.

The

grid

sear

chis

equa

llyfa

stfo

rre

ctan

gula

ran

dtr

iclin

icbo

xes.

Thu

sfo

rm

ost

prot

ein

and

pept

ide

sim

ulat

ions

the

rhom

bic

dode

cahe

dron

will

beth

epr

efer

able

box

shap

e.

Cha

rge

grou

ps

Whe

reap

plic

able

,nei

ghbo

rsea

rchi

ngis

carr

ied

outo

nth

eba

sis

ofch

arg

eg

rou

ps.A

char

gegr

oup

isa

smal

lset

ofne

arby

atom

sth

atha

vene

tcha

rge

zero

.C

harg

egr

oups

are

defin

edin

the

mol

ecul

arto

polo

gy.

Ifth

ene

ares

tim

age

dist

ance

betw

een

the

geo

me

tric

alc

en

terso

fthe

atom

sof

two

char

gegr

oups

isle

ssth

anth

ecu

toff

radi

us,a

llat

ompa

irsbe

twee

nth

ech

arge

grou

psar

ein

clud

edin

the

pair

list.

Thi

spr

oced

ure

avoi

dsth

ecr

eatio

nof

char

ges

due

toth

eus

eof

acu

t-of

f(w

hen

one

char

geof

adi

pole

isw

ithin

rang

ean

dth

eot

her

not)

,whi

chca

nha

vedi

sast

rous

cons

eque

nces

for

the

beha

vior

ofth

eC

oulo

mb

inte

ract

ion

func

tion

atdi

stan

ces

near

the

cut-

offr

adiu

s.If

mol

ecul

argr

oups

have

full

char

ges

(ions

),ch

arge

grou

psdo

nota

void

adve

rse

cut-

offe

ffect

s,an

dyo

ush

ould

cons

ider

usin

gon

eof

the

latti

cesu

mm

etho

dssu

pplie

dby

GR

OM

AC

S[

13].

Ifap

prop

riate

lyco

nstr

ucte

dsh

iftfu

nctio

nsar

eus

edfo

rth

eel

ectr

osta

ticfo

rces

,no

char

gegr

oups

Page 40: CS USER AL Simulations - GROMACS

20

Ch

ap

ter

3.

Alg

orith

ms

areneeded.

Such

shiftfunctionsare

implem

entedin

GR

OM

AC

S(see

chapter4)

butmustbe

usedw

ithcare:

inprinciple

theyshould

becom

binedw

itha

latticesum

forlong-range

electrostatics.

3.4.3C

ompute

forces

Potentialenergy

When

forcesare

computed,

thepotential

energyof

eachinteraction

termis

computed

asw

ell.T

hetotalpotentialenergy

issum

med

forvarious

contributions,suchas

Lennard-Jones,Coulom

b,and

bondedterm

s.Itis

alsopossible

tocom

putethese

contributionsfor

gro

up

sofatoms

thatareseparately

defined(see

sec.3.3).

Kinetic

energyand

temperature

The

temperature

isgiven

bythe

totalkineticenergy

oftheN

-particlesystem

:

Ekin

=12

N∑i=1

mi v

2i(3.11)

From

thisthe

absolutetem

peratureT

canbe

computed

using:

12N

df kT

=E

kin

(3.12)

where

kis

Boltzm

ann’sconstant

andNdf

isthe

number

ofdegrees

offreedom

which

canbe

computed

from:

Ndf

=3N

−N

c −N

com

(3.13)

Here

Nc

isthe

numberofco

nstra

intsim

posedon

thesystem

.W

henperform

ingm

oleculardynam-

icsN

com

=3

additionaldegreesof

freedomm

ustbe

removed,

becausethe

threecenter-of-m

assvelocities

areconstants

ofthe

motion,

which

areusually

setto

zero.W

hensim

ulatingin

vacuo,the

rotationaround

thecenter

ofm

asscan

alsobe

removed,

inthis

caseN

com

=6.

When

more

thanone

temperature

couplinggroup

isused,the

number

ofdegreesoffreedom

forgroup

iis:

Nidf

=(3N

i−N

ic ) 3N−N

c −N

com

3N−N

c(3.14)

The

kineticenergy

canalso

bew

rittenas

atensor,w

hichis

necessaryfor

pressurecalculation

ina

triclinicsystem

,orsystem

sw

hereshear

forcesare

imposed:

Ekin

=12

N∑i

mi v

i ⊗v

i(3.15)

Pressure

andvirial

The

pressuretensorPis

calculatedfrom

thedifference

between

kineticenergy

Ekin

andthe

virialΞ

P=

2V(E

kin−

Ξ)

(3.16)

E.4

1.

gro

tacf

20

9

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-n

ind

ex.n

dx

Input,Opt.

Indexfile

-qe

iwit.p

db

Input,Opt.

Protein

databank

file-o

qb

fac.p

db

Output,O

pt.P

roteindata

bankfile

-ox

xave

r.pd

bO

utput,Opt.

Protein

databank

file-o

rmsf.xvg

Output

xvgr/xmgr

file-o

drm

sde

v.xvgO

utput,Opt.

xvgr/xmgr

file-o

cco

rrel.xvg

Output,O

pt.xvgr/xm

grfile

-dir

rmsf.lo

gO

utput,Opt.

Logfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-re

sbool

no

Calculate

averagesfor

eachresidue

-an

isobool

no

Com

puteanisotropic

termperature

factors

E.41

grotacf

grotacf

calculatesthe

rotationalcorrelationfunction

form

olecules.T

hreeatom

s(i,j,k)

must

begiven

inthe

indexfile,defining

two

vectorsijand

jk.T

herotationalacfis

calculatedas

theautocorrelation

functionofthe

vectorn

=ijx

jk,i.e.the

crossproductofthe

two

vectors.S

incethree

atoms

spana

plane,theorder

ofthethree

atoms

doesnotm

atter.O

ptionally,controlledby

the-d

switch,you

cancalculate

therotational

correlationfunction

forlinear

molecules

byspecifying

two

atoms

(i,j)in

theindex

file.

EX

AM

PLE

S

grotacf-P

1-nparm

2-fft-n

index-o

rotacf-x-P1

-faexpfit-x-P

1-beginfit2.5

-endfit20.0

This

willcalculate

therotationalcorrelation

functionusing

afirst

orderLegendre

polynomialof

theangle

ofavector

definedby

theindex

file.T

hecorrelation

functionw

illbefitted

from2.5

pstill20.0

psto

atw

oparam

eterexponential

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-nin

de

x.nd

xInput

Indexfile

-oro

tacf.xvg

Output

xvgr/xmgr

file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-d

booln

oU

seindex

doublets(vectors)

forcorrelation

functioninstead

oftriplets

(planes)-a

ver

boolye

sA

verageover

molecules

Page 41: CS USER AL Simulations - GROMACS

20

8A

pp

en

dix

E.

Ma

nu

alP

age

s

0.6

inth

isca

se)

can

bege

nera

ted,

byde

faul

tave

ragi

ngov

ereq

uiva

lent

hydr

ogen

s(a

lltr

iple

tsof

hydr

ogen

sna

med

*[12

3]).

Add

ition

ally

alis

tof

equi

vale

ntat

oms

can

besu

pplie

d(

-eq

uiv

),ea

chlin

eco

ntai

ning

ase

tofe

quiv

alen

tato

ms

spec

ified

asre

sidu

enu

mbe

ran

dna

me

and

atom

nam

e;e.

g.:

3S

ER

HB

13

SE

RH

B2

Res

idue

and

atom

nam

esm

ust

exac

tlym

atch

thos

ein

the

stru

ctur

efil

e,in

clud

ing

case

.S

peci

fyin

gno

n-se

quen

tiala

tom

sis

unde

fined

.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Str

uctu

re+

mas

s(db

):tp

rtp

btp

agr

og9

6pd

b-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-eq

uiv

eq

uiv

.da

tIn

put,

Opt

.G

ener

icda

tafil

e-o

dis

trm

sd.x

vgO

utpu

txv

gr/x

mgr

file

-rm

srm

sdis

t.xp

mO

utpu

t,O

pt.

XP

ixM

apco

mpa

tible

mat

rixfil

e-s

clrm

ssca

le.x

pm

Out

put,

Opt

.X

Pix

Map

com

patib

lem

atrix

file

-me

an

rmsm

ea

n.x

pm

Out

put,

Opt

.X

Pix

Map

com

patib

lem

atrix

file

-nm

r3n

mr3

.xp

mO

utpu

t,O

pt.

XP

ixM

apco

mpa

tible

mat

rixfil

e-n

mr6

nm

r6.x

pm

Out

put,

Opt

.X

Pix

Map

com

patib

lem

atrix

file

-no

en

oe

.da

tO

utpu

t,O

pt.

Gen

eric

data

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-n

leve

lsin

t4

0D

iscr

etiz

erm

sin

#le

vels

-ma

xre

al-1

Max

imum

leve

lin

mat

rices

-su

mh

bool

yes

aver

age

dist

ance

over

equi

vale

nthy

drog

ens

E.4

0g

rmsf

grm

sfco

mpu

tes

the

root

mea

nsq

uare

fluct

uatio

n(R

MS

F,i.e

.st

anda

rdde

viat

ion)

ofat

omic

posi

tions

afte

rfir

stfit

ting

toa

refe

renc

efr

ame.

With

optio

n-o

qth

eR

MS

Fva

lues

are

conv

erte

dto

B-f

acto

rva

lues

,whi

char

ew

ritte

nto

apd

bfil

ew

ithth

eco

ordi

nate

s,of

the

stru

ctur

efil

e,or

ofa

pdb

file

whe

n-q

issp

ecifi

ed.

Opt

ion-o

xw

rites

the

B-f

acto

rsto

afil

ew

ithth

eav

erag

eco

ordi

nate

s.

With

the

optio

n-o

dth

ero

otm

ean

squa

rede

viat

ion

with

resp

ectt

oth

ere

fere

nce

stru

ctur

eis

calc

ulat

ed.

With

the

optio

nan

iso

grm

sfw

illco

mpu

tean

isot

ropi

cte

mpe

ratu

refa

ctor

san

dth

enit

will

also

outp

utav

erag

eco

ordi

nate

san

da

pdb

file

with

AN

ISO

Ure

cord

s(c

orre

sond

ing

toth

e-o

qor

-ox

optio

n).

Ple

ase

note

that

the

Uva

lues

are

orie

ntat

ion

depe

nden

t,so

befo

reco

mpa

rison

with

expe

rimen

tald

ata

you

shou

ldve

rify

that

you

fitto

the

expe

rimen

talc

oord

inat

es.

Whe

na

pdb

inpu

tfile

ispa

ssed

toth

epr

ogra

man

dth

e-a

nis

ofla

gis

seta

corr

elat

ion

plot

ofth

eU

ijw

illbe

crea

ted,

ifan

yan

isot

ropi

cte

mpe

ratu

refa

ctor

sar

epr

esen

tin

the

pdb

file.

With

optio

n-d

irth

eav

erag

eM

SF

(3x3

)m

atrix

isdi

agon

aliz

ed.

Thi

ssh

ows

the

dire

ctio

nsin

whi

chth

eat

oms

fluct

uate

the

mos

tand

the

leas

t.

File

s

3.4

.M

ole

cula

rD

yna

mic

s2

1

12

0t

xv

x

Fig

ure

3.6:

The

Leap

-Fro

gin

tegr

atio

nm

etho

d.T

heal

gorit

hmis

calle

dLe

ap-F

rog

beca

use

ran

dv

are

leap

ing

like

frog

sov

erea

chot

hers

back

.

whe

reV

isth

evo

lum

eof

the

com

puta

tiona

lbox

.T

hesc

alar

pres

sure

P,

whi

chca

nbe

used

for

pres

sure

coup

ling

inth

eca

seof

isot

ropi

csy

stem

s,is

com

pute

das

:

P=

trac

e(P

)/3

(3.1

7)

The

viria

lΞte

nsor

isde

fined

as

Ξ=−

1 2

∑ i<j

rij⊗

Fij

(3.1

8)

Inse

c.B

.1th

eim

plem

enta

tion

inG

RO

MA

CS

ofth

evi

rialc

ompu

tatio

nis

desc

ribed

.

3.4.

4U

pdat

eco

nfigu

ratio

n

The

GR

OM

AC

SM

Dpr

ogra

mut

ilize

sth

eso

-cal

ledlea

p-f

rog

algo

rithm

[14]

for

the

inte

grat

ion

ofth

eeq

uatio

nsof

mot

ion.

The

leap

-fro

gal

gorit

hmus

espo

sitio

nsr

attim

et

and

velo

citie

svat

timet−

∆t

2;i

tupd

ates

posi

tions

and

velo

citie

sus

ing

the

forc

esF

(t)

dete

rmin

edby

the

posi

tions

attim

et:

v(t

+∆t 2)

=v(t−

∆t 2)+

F(t

)m

∆t

(3.1

9)

r(t

+∆t)

=r(t

)+

v(t

+∆t 2)∆t

(3.2

0)

The

algo

rithm

isvi

sual

ized

inF

ig.3.6.

Itis

equi

vale

ntto

the

Verle

t[15]

algo

rithm

:

r(t

+∆t)

=2r

(t)−

r(t−

∆t)

+F

(t)

m∆t2

+O

(∆t4

)(3

.21)

The

algo

rithm

isof

third

orde

rinr

and

istim

e-re

vers

ible

.S

eere

f.[

16]

for

the

mer

itsof

this

algo

rithm

and

com

paris

onw

ithot

her

time

inte

grat

ion

algo

rithm

s.

The

equa

tions

ofm

otio

nar

em

odifi

edfo

rte

mpe

ratu

reco

uplin

gan

dpr

essu

reco

uplin

g,an

dex

-te

nded

toin

clud

eth

eco

nser

vatio

nof

cons

trai

nts,

allo

fwhi

char

ede

scrib

edbe

low

.

3.4.

5Te

mpe

ratu

reco

uplin

g

For

seve

ralr

easo

ns(d

riftd

urin

geq

uilib

ratio

n,dr

iftas

are

sult

offo

rce

trun

catio

nan

din

tegr

atio

ner

rors

,hea

ting

due

toex

tern

alor

fric

tiona

lfor

ces)

,iti

sne

cess

ary

toco

ntro

lthe

tem

pera

ture

ofth

esy

stem

.G

RO

MA

CS

can

use

eith

erth

ew

ea

kco

up

lings

chem

eof

Ber

ends

en[

17]o

rth

eex

tend

eden

sem

ble

Nose

-Hoo

ver

sche

me

[18,1

9].

Page 42: CS USER AL Simulations - GROMACS

22

Ch

ap

ter

3.

Alg

orith

ms

Berendsen

temperature

coupling

The

Berendsen

algorithmm

imics

weak

couplingw

ithfirst-order

kineticsto

anexternalheatbath

with

giventem

peratureT0 .

See

ref.[20]

fora

comparison

with

theN

ose-Hoover

scheme.

The

effectof

thisalgorithm

isthat

adeviation

ofthe

systemtem

peraturefromT

0is

slowly

correctedaccording

todTdt

=T

0 −T

τ(3.22)

which

means

thata

temperature

deviationdecays

exponentiallyw

itha

time

constantτ.

This

method

ofcouplinghas

theadvantage

thatthestrength

ofthecoupling

canbe

variedand

adaptedto

theuser

requirement:

forequilibration

purposesthe

couplingtim

ecan

betaken

quiteshort(e.g.

0.01ps),butfor

reliableequilibrium

runsitcan

betaken

much

longer(e.g.0.5

ps)in

which

caseithardly

influencesthe

conservativedynam

ics.

The

heatflowinto

oroutofthe

systemis

effectedby

scalingthe

velocitiesofeach

particleevery

stepw

itha

time-dependentfactor

λ,given

by

λ= [1

+∆t

τT {

T0

T(t−

∆t

2)−

1 }]1/2

(3.23)

The

parameterτ

Tis

closeto,

butnot

exactlyequal

tothe

time

constantτ

ofthe

temperature

coupling(eqn.3.22):

τ=

2CVτT/N

df k(3.24)

where

CV

isthe

totalheat

capacityof

thesystem

,k

isB

oltzmann’s

constant,andN

dfis

thetotal

number

ofdegrees

offreedom

.T

hereason

thatτ6=τT

isthat

thekinetic

energychange

causedby

scalingthe

velocitiesis

partlyredistributed

between

kineticand

potentialenergyand

hencethe

changein

temperature

isless

thanthe

scalingenergy.

Inpractice,the

ratioτ/τ

Tranges

from1

(gas)to

2(harm

onicsolid)

to3

(water).

When

we

usethe

term’tem

peraturecoupling

time

constant’,w

em

eanthe

parameterτT

.N

otethat

inpractice

thescaling

factorλislim

itedto

therange

of0.8<

=λ<

=1.25,

toavoid

scalingby

verylarge

numbers

which

may

crashthe

simulation.

Innorm

aluse,λwillalw

aysbe

much

closerto

1.0.

Strictly,forcom

putingthe

scalingfactorthe

temperature

Tis

neededattim

et,butthisis

notavail-able

inthe

algorithm.

Inpractice,

thetem

peratureat

theprevious

time

stepis

used(as

indicatedin

eqn.3.23),which

isperfectly

allrightsincethe

couplingtim

econstantis

much

longerthan

onetim

estep.

The

Berendsen

algorithmis

stableup

toτT≈

∆t.

Nos

e-Hoover

temperature

coupling

The

Berendsen

weak

couplingalgorithm

isextrem

elyefficientfor

relaxinga

systemto

thetarget

temperature,

butonce

yoursystem

hasreached

equilibriumit

might

bem

oreim

portantto

probea

correctcanonicalensem

ble.T

hisis

unfortunatelynot

thecase

forthe

weak

couplingschem

e,although

thedifference

isusually

negligible.

Toenable

canonicalensemble

simulations,

GR

OM

AC

Salso

supportsthe

extended-ensemble

ap-proach

firstproposedby

Nose[18]and

laterm

odifiedby

Hoover[19].

The

systemH

amiltonian

isextended

byintroducing

atherm

alreservoirand

afriction

termin

theequations

ofm

otion.T

he

E.3

9.

grm

sdist

20

7

Allthe

structuresare

fittedpairw

ise.

With

-f2,the

’otherstructures’are

takenfrom

asecond

trajectory.

Option

-bin

doesa

binarydum

pofthe

comparison

matrix.

Option

-bm

producesa

matrix

ofaveragebond

angledeviations

analogouslyto

the-m

option.O

nlybonds

between

atoms

inthe

comparison

groupare

considered.

Files

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-f

traj.xtc

InputG

enerictrajectory:

xtctrr

trjgrog96

pdb-f2

traj.xtc

Input,Opt.

Generic

trajectory:xtc

trrtrjgro

g96pdb

-nin

de

x.nd

xInput,O

pt.Index

file-o

rmsd

.xvgO

utputxvgr/xm

grfile

-mir

rmsd

mir.xvg

Output,O

pt.xvgr/xm

grfile

-aa

vgrp

.xvgO

utput,Opt.

xvgr/xmgr

file-d

istrm

sd-d

ist.xvgO

utput,Opt.

xvgr/xmgr

file-m

rmsd

.xpm

Output,O

pt.X

PixM

apcom

patiblem

atrixfile

-bin

rmsd

.da

tO

utput,Opt.

Generic

datafile

-bm

bo

nd

.xpm

Output,O

pt.X

PixM

apcom

patiblem

atrixfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-tuenum

ps

Tim

eunit:p

s,fs

,ns

,us

,ms,s

,morh

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-w

ha

tenum

rmsd

Structuraldifference

measure:

rmsd

,rho

orrho

sc-p

bc

boolye

sP

BC

check-fit

boolye

sF

ittoreference

structure-p

rev

int0

Com

parew

ithprevious

frame

-split

booln

oS

plitgraphw

heretim

eis

zero-skip

int1

Only

write

everynr-th

frame

tom

atrix-skip

2int

1O

nlyw

riteevery

nr-thfram

eto

matrix

-ma

xreal

-1M

aximum

levelincom

parisonm

atrix-m

inreal

-1M

inimum

levelincom

parisonm

atrix-b

ma

xreal

-1M

aximum

levelinbond

anglem

atrix-b

min

real-1

Minim

umlevelin

bondangle

matrix

-nle

vels

int8

0N

umber

oflevelsin

them

atrices

E.39

grm

sdist

grm

sdistcom

putesthe

rootm

eansquare

deviationof

atomdistances,

which

hasthe

advantagethat

nofit

isneeded

likein

standardR

MS

deviationas

computed

byg

rms.

The

referencestructure

istaken

fromthe

structurefile.

The

rmsd

attime

tiscalculated

asthe

rms

ofthedifferences

indistance

between

atom-pairs

inthe

referencestructure

andthe

structureattim

et.

grm

sdistcanalso

producem

atricesofthe

rms

distances,rms

distancesscaled

with

them

eandistance

andthe

mean

distancesand

matrices

with

NM

Raveraged

distances(1/r

3and

1/r 6averaging).

Finally,

listsof

atompairs

with

1/r 3and1/r 6

averageddistance

belowthe

maxim

umdistance

(-m

ax

,which

willdefaultto

Page 43: CS USER AL Simulations - GROMACS

20

6A

pp

en

dix

E.

Ma

nu

alP

age

s

Ifa

run

inpu

tfil

eis

supp

lied

(-s),

excl

usio

nsde

fined

inth

atfil

ear

eta

ken

into

acco

unt

whe

nca

lcul

atin

gth

erd

f.T

heop

tion-

cut

ism

eant

asan

alte

rnat

ive

way

toav

oid

intr

amol

ecul

arpe

aks

inth

erd

fplo

t.It

isho

wev

erbe

tter

tosu

pply

aru

nin

put

file

with

ahi

gher

num

ber

ofex

clus

ions

.F

oreg

.be

nzen

ea

topo

logy

with

nrex

clse

tto

5w

ould

elim

inat

eal

lint

ram

olec

ular

cont

ribut

ions

toth

erd

f.N

ote

that

alla

tom

sin

the

sele

cted

grou

psar

eus

ed,a

lso

the

ones

that

don’

thav

eLe

nnar

d-Jo

nes

inte

ract

ions

.

Opt

ion

-cn

prod

uces

the

cum

ulat

ive

num

ber

rdf.

Tobr

idge

the

gap

betw

een

theo

ryan

dex

perim

ent

stru

ctur

efa

ctor

sca

nbe

com

pute

d(o

ptio

n-s

q).

The

algo

rithm

uses

FF

T,th

egr

idsp

acin

gof

whi

chis

dete

rmin

edby

optio

n-g

rid

.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put,

Opt

.S

truc

ture

+m

ass(

db):

tpr

tpb

tpa

gro

g96

pdb

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e-o

rdf.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e-s

qsq

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-cn

rdf

cn.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-h

qh

q.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-im

ag

esq

.xp

mO

utpu

t,O

pt.

XP

ixM

apco

mpa

tible

mat

rixfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-b

inre

al0

.00

1B

inw

idth

(nm

)-c

om

bool

no

RD

Fw

ithre

spec

tto

the

cent

erof

mas

sof

first

grou

p-c

ut

real

0S

hort

estd

ista

nce

(nm

)to

beco

nsid

ered

-fa

de

real

0F

rom

this

dist

ance

onw

ards

the

RD

Fis

tran

form

edby

g’(r

)=

1+

[g(r

)-1]

exp(

-(r/

fade

-1)2

tom

ake

itgo

to1

smoo

thly

.If

fade

is0.

0no

thin

gis

done

.-g

rid

real

0.0

5G

ridsp

acin

g(in

nm)

for

FF

Ts

whe

nco

mpu

ting

stru

ctur

efa

ctor

s-n

leve

lin

t2

0N

umbe

rof

diffe

rent

colo

rsin

the

diffr

actio

nim

age

-wa

vere

al0

.1W

avel

engt

hfo

rX

-ray

s/N

eutr

ons

for

scat

terin

g.0.

1nm

corr

espo

nds

toro

ughl

y12

keV

E.3

8g

rms

grm

sco

mpa

res

two

stru

ctur

esby

com

putin

gth

ero

otm

ean

squa

rede

viat

ion

(RM

SD

),th

esi

ze-in

depe

nden

t’rh

o’si

mila

rity

para

met

er(r

ho)

orth

esc

aled

rho

(rho

sc),

refe

renc

eM

aior

ov&

Crip

pen,

PR

OT

EIN

S22

,27

3(1

995)

.T

his

isse

lect

edby-w

ha

t.

Eac

hst

ruct

ure

from

atr

ajec

tory

(-f

)is

com

pare

dto

are

fere

nce

stru

ctur

efr

oma

run

inpu

tfil

eby

leas

t-sq

uare

sfit

ting

the

stru

ctur

eson

top

ofea

chot

her.

The

refe

renc

est

ruct

ure

ista

ken

from

the

stru

ctur

efil

e(-

s).

With

optio

n-m

iral

soa

com

paris

onw

ithth

em

irror

imag

eof

the

refe

renc

est

ruct

ure

isca

lcul

ated

.

Opt

ion

-pre

vpr

oduc

esth

eco

mpa

rison

with

apr

evio

usfr

ame.

Opt

ion

-mpr

oduc

esa

mat

rixin.

xpm

form

atof

com

paris

onva

lues

ofea

chst

ruct

ure

inth

etr

ajec

tory

with

resp

ect

toea

chot

her

stru

ctur

e.T

his

file

can

bevi

sual

ized

with

for

inst

ance

xvan

dca

nbe

conv

erte

dto

post

scrip

twith

xpm

2p

s.

3.4

.M

ole

cula

rD

yna

mic

s2

3

fric

tion

forc

eis

prop

ortio

nalt

oth

epr

oduc

tof

each

part

icle

’sve

loci

tyan

da

fric

tion

para

met

erξ.

Thi

sfr

ictio

npa

ram

eter

(or

’hea

tbat

h’va

riabl

e)is

afu

llydy

nam

icqu

antit

yw

ithits

own

equa

tion

ofm

otio

n;th

etim

ede

rivat

ive

isca

lcul

ated

from

the

diffe

renc

ebe

twee

nth

ecu

rren

tkin

etic

ener

gyan

dth

ere

fere

nce

tem

pera

ture

.

Inth

efo

rmul

atio

nof

Hoo

ver,

the

part

icle

s’eq

uatio

nsof

mot

ion

inF

ig.

3.3

are

repl

aced

by

d2r

i

dt2

=F

i

mi−ξ

dri

dt,

(3.2

5)

whe

reth

eeq

uatio

nof

mot

ion

for

the

heat

bath

para

met

erξ

is

dξ dt=

1 Q(T

−T

0).

(3.2

6)

The

refe

renc

ete

mpe

ratu

reis

deno

ted

T0,

whi

leT

isth

ecu

rren

tm

omen

tary

tem

pera

ture

ofth

esy

stem

.T

hest

reng

thof

the

coup

ling

isde

term

ined

byth

eco

nsta

ntQ

(usu

ally

calle

dth

e’m

ass

para

met

er’o

fthe

rese

rvoi

r)in

com

bina

tion

with

the

refe

renc

ete

mpe

ratu

re.

Inou

rop

inio

n,th

em

ass

para

met

eris

aso

mew

hata

wkw

ard

way

ofde

scrib

ing

coup

ling

stre

ngth

,es

peci

ally

due

toits

depe

nden

ceon

refe

renc

ete

mpe

ratu

re(a

ndso

me

impl

emen

tatio

nsev

enin

-cl

ude

the

num

ber

ofde

gree

sof

free

dom

inyo

ursy

stem

whe

nde

finin

gQ

).To

mai

ntai

nth

eco

u-pl

ing

stre

ngth

,you

wou

ldha

veto

chan

geQpr

opor

tiona

lto

your

chan

gein

refe

renc

ete

mpe

ratu

re.

For

this

reas

on,w

epr

efer

tole

tthe

GR

OM

AC

Sus

erw

ork

with

the

perio

dτ T

ofth

eos

cilla

tions

ofki

netic

ener

gybe

twee

nth

esy

stem

and

the

rese

rvoi

rin

stea

d.It

isdi

rect

lyre

late

dto

Qan

dT

0as

Q=τ

2 TT

0

4π2.

(3.2

7)

Thi

spr

ovid

esa

muc

hm

ore

intu

itive

way

ofse

lect

ing

the

Nos

e-H

oove

rco

uplin

gst

reng

th(s

imila

rto

the

wea

kco

uplin

gre

laxa

tion)

,an

din

addi

tionτ T

isin

depe

nden

tof

syst

emsi

zean

dre

fere

nce

tem

pera

ture

.

Itis

how

ever

impo

rtan

tto

keep

the

diffe

renc

ebe

twee

nth

ew

eak

coup

ling

sche

me

and

the

Nos

e-H

oove

ral

gorit

hmin

min

d:U

sing

wea

kco

uplin

gyo

uge

ta

stro

ngly

dam

ped

exp

on

en

tialre

lax-

atio

n,w

hile

the

Nos

e-H

oove

rap

proa

chpr

oduc

esanos

cilla

tory

rela

xatio

n.T

heac

tual

time

itta

kes

tore

lax

with

Nose

-Hoo

ver

coup

ling

isse

vera

ltim

esla

rger

than

the

perio

dof

the

osci

llatio

nsth

atyo

use

lect

.T

hese

osci

llatio

ns(in

cont

rast

toex

pone

ntia

lrel

axat

ion)

also

mea

nsth

atth

etim

eco

nsta

ntno

rmal

lysh

ould

be4–

5tim

esla

rger

than

the

rela

xatio

ntim

eus

edw

ithw

eak

coup

ling,

buty

our

mile

age

may

vary

.

3.4.

6P

ress

ure

coup

ling

Inth

esa

me

spiri

tas

the

tem

pera

ture

coup

ling,

the

syst

emca

nal

sobe

coup

led

toa

’pre

ssur

eba

th’.

GR

OM

AC

Ssu

ppor

tsbo

thth

eB

eren

dsen

algo

rithm

[17

]th

atsc

ales

coor

dina

tes

and

box

vect

ors

ever

yst

ep,a

ndth

eex

tend

eden

sem

ble

Par

rinel

lo-R

ahm

anap

proa

ch.

Bot

hof

thes

eca

nbe

com

bine

dw

ithan

yof

the

tem

pera

ture

coup

ling

met

hods

abov

e.

Page 44: CS USER AL Simulations - GROMACS

24

Ch

ap

ter

3.

Alg

orith

ms

Berendsen

pressurecoupling

The

Berendsen

algorithmrescales

thecoordinates

andbox

vectorsevery

stepw

itha

matrix

µ,

which

hasthe

effectof

afirst-order

kineticrelaxation

ofthe

pressuretow

ardsa

givenreference

pressureP0 :

dPdt=

P0 −

Pτp

(3.28)

The

scalingm

atrixµis

givenbyµ

ij=δij −

∆t

3τpβ

ij {P

0ij −

Pij (t)}

(3.29)

Here

βis

theisotherm

alcom

pressibilityof

thesystem

.In

most

casesthis

will

bea

diagonalm

atrix,with

equalelements

onthe

diagonal,thevalue

ofwhich

isgenerally

notknown.

Itsufficesto

takea

roughestim

atebecause

thevalue

ofβ

onlyinfluences

thenon-criticaltim

econstant

ofthe

pressurerelaxation

withoutaffecting

theaverage

pressureitself.

For

water

at1atm

and300

=4.6

×10−

10

Pa −

1=

4.6×

10−

5B

ar −1,w

hichis7.6

×10−

4M

Dunits

(seechapter2).

Most

otherliquids

havesim

ilarvalues.

When

scalingcom

pletelyanisotropically,

thesystem

hasto

berotated

inorder

toobey

thebox

restriction(

3.1).T

hisrotation

isapproxim

atedin

firstorderin

thescaling,w

hichis

usuallyless

than10−

4.T

heactualscaling

matrixµ

′is:

µ′=

µxx

µxy

yx

µxz+µ

zx

yy

µyz+µ

zy

00

µzz

(3.30)

The

velocitiesare

neitherscaled

norrotated.

InG

RO

MA

CS

,the

Berendsen

scalingcan

alsobe

doneisotropically,

which

means

thatinstead

ofP

adiagonalm

atrixw

ithelem

entsof

sizetrace(P

)/3is

used.F

orsystem

sw

ithinterfaces,

semi-isotropic

scalingcan

beuseful.

Inthis

casethex/y-directions

arescaled

isotropicallyand

thez

directionis

scaledindependently.

The

compressibility

inthex/y

orz-direction

canbe

settozero,to

scaleonly

inthe

otherdirection(s).

Ifyou

allowfullanisotropic

deformations

anduse

constraintsyou

might

haveto

scaleslow

eror

decreaseyour

timestep

toavoid

errorsfrom

theconstraintalgorithm

s.

Parrinello-R

ahman

pressurecoupling

Incases

where

thefluctuations

inpressure

orvolum

eare

important

pe

rse(e.g.

tocalculate

ther-m

odynamic

properties)it

might

atleast

theoreticallybe

aproblem

thatthe

exactensem

bleis

notw

ell-definedfor

thew

eakcoupling

scheme.

Forthis

reason,GR

OM

AC

Salso

supportsconstant-pressure

simulations

usingthe

Parrinello-R

ahman

approach[21,22],which

issim

ilarto

theN

ose-Hoover

temperature

coupling.W

iththe

Parrinello-

Rahm

anbarostat,

thebox

vectorsas

representedby

them

atrixb

obeythe

matrix

equationof

motion 1

1The

boxm

atrixrepresentationbin

GR

OM

AC

Scorresponds

tothe

transposeofthe

boxm

atrixrepresentation

hin

thepaper

byN

oseand

Klein.

Because

ofthis,some

ofourequations

willlook

slightlydifferent.

E.3

6.

gra

ma

20

5

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-nin

de

x.nd

xInput

Indexfile

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-op

ote

ntia

l.xvgO

utputxvgr/xm

grfile

-oc

cha

rge

.xvgO

utputxvgr/xm

grfile

-of

field

.xvgO

utputxvgr/xm

grfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-d

stringZ

Takethe

normalon

them

embrane

indirection

X,Y

orZ

.-sl

int1

0C

alculatepotential

asfunction

ofboxlength,

dividingthe

boxin

#nrslices.

-cbint

0D

iscardfirst#nr

slicesofbox

forintegration

-ceint

0D

iscardlast#nr

slicesofbox

forintegration

-tzreal

0T

ranslateallcoordinates<distance>

inthe

directionofthe

box-sp

he

rical

booln

oC

alculatesphericalthingie

•D

iscardingslices

forintegration

shouldnotbe

necessary.

E.36

gram

a

gram

aselects

theP

hi/Psidihedralcom

binationsfrom

yourtopology

fileand

computes

theseas

afunction

oftime.

Using

simple

Unix

toolssuch

asgrep

youcan

selectoutspecificresidues.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-ora

ma

.xvgO

utputxvgr/xm

grfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files

E.37

grdf

The

structureof

liquidscan

bestudied

byeither

neutronor

X-ray

scattering.T

hem

ostcom

mon

way

todescribe

liquidstructure

isby

aradial

distributionfunction.

How

ever,this

isnot

easyto

obtainfrom

ascattering

experiment.

grdf

calculatesradial

distributionfunctions

indifferent

ways.

The

normal

method

isaround

a(set

of)particle(s),the

otherm

ethodis

aroundthe

centerofm

assofa

setofparticles.

Page 45: CS USER AL Simulations - GROMACS

20

4A

pp

en

dix

E.

Ma

nu

alP

age

s

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e-o

en

sem

ble

.xtc

Out

put

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-te

mp

real

30

0Te

mpe

ratu

rein

Kel

vin

-se

ed

int

-1R

ando

mse

ed,-

1ge

nera

tes

ase

edfr

omtim

ean

dpi

d-n

um

int

10

0N

umbe

rof

stru

ctur

esto

gene

rate

-first

int

7F

irste

igen

vect

orto

use

(-1

isse

lect

)-la

stin

t-1

Last

eige

nvec

tor

tous

e(-

1is

tillt

hela

st)

E.3

4g

orde

r

Com

pute

the

orde

rpa

ram

eter

per

atom

for

carb

onta

ils.

For

atom

ithe

vect

ori-1

,i+

1is

used

toge

ther

with

anax

is.

The

inde

xfil

eha

sto

cont

ain

agr

oup

with

alle

quiv

alen

tato

ms

inal

ltai

lsfo

rea

chat

omth

eor

der

para

met

erha

sto

beca

lcul

ated

for.

The

prog

ram

can

also

give

alld

iago

nale

lem

ents

ofth

eor

der

tens

oran

dev

enca

lcul

ate

the

deut

eriu

mor

der

para

met

erS

cd(d

efau

lt).

Ifth

eop

tion

-szo

nly

isgi

ven,

only

one

orde

rte

nsor

com

pone

nt(s

peci

fied

byth

e-d

optio

n)is

give

nan

dth

eor

der

para

met

erpe

rsl

ice

isca

lcul

ated

asw

ell.

If-s

zonl

yis

nots

elec

ted,

alld

iago

nale

lem

ents

and

the

deut

eriu

mor

der

para

met

eris

give

n.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-nin

de

x.n

dx

Inpu

tIn

dex

file

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a-o

ord

er.

xvg

Out

put

xvgr

/xm

grfil

e-o

dd

eu

ter.

xvg

Out

put

xvgr

/xm

grfil

e-o

ssl

ice

d.x

vgO

utpu

txv

gr/x

mgr

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-d

enum

zD

irect

ion

ofth

eno

rmal

onth

em

embr

ane:z,x

ory

-sl

int

1C

alcu

late

orde

rpa

ram

eter

asfu

nctio

nof

boxl

engt

h,di

vidi

ngth

ebo

xin

#nr

slic

es.

-szo

nly

bool

no

Onl

ygi

veS

zel

emen

tofo

rder

tens

or.

(axi

sca

nbe

spec

ified

with

-d)

-un

sat

bool

no

Cal

cula

teor

derp

aram

eter

sfo

runs

atur

ated

carb

ons.

Not

eth

atth

isca

nnot

bem

ixed

with

norm

alor

der

para

met

ers.

E.3

5g

pote

ntia

l

Com

pute

the

elec

tros

tatic

alpo

tent

iala

cros

sth

ebo

x.T

hepo

tent

iali

scal

cula

ted

byfir

stsu

mm

ing

the

char

ges

per

slic

ean

dth

enin

tegr

atin

gtw

ice

ofth

isch

arge

dist

ribut

ion.

Per

iodi

cbo

unda

ries

are

not

take

nin

toac

-co

unt.

Ref

eren

ceof

pote

ntia

lis

take

nto

beth

ele

ftsi

deof

the

box.

It’s

also

poss

ible

toca

lcul

ate

the

pote

ntia

lin

sphe

rical

coor

dina

tes

asfu

nctio

nof

rby

calc

ulat

ing

ach

arge

dist

ribut

ion

insp

heric

alsl

ices

and

twic

ein

tegr

atin

gth

em.

epsi

lonris

take

nas

1,2

ism

ore

appr

opria

tein

man

yca

ses

3.4

.M

ole

cula

rD

yna

mic

s2

5

db2

dt2

=V

W−

1b′−

1(P

−P

ref

).(3

.31)

The

volu

me

ofth

ebo

xis

deno

tedV,a

ndW

isa

mat

rixpa

ram

eter

that

dete

rmin

esth

est

reng

thof

the

coup

ling.

The

mat

ricesP

andP

ref

are

the

curr

enta

ndre

fere

nce

pres

sure

s,re

spec

tivel

y.

The

equa

tions

ofm

otio

nfo

rth

epa

rtic

les

are

also

chan

ged,

just

asfo

rth

eN

ose-

Hoo

ver

coup

ling.

Inm

ostc

ases

you

wou

ldco

mbi

neth

eP

arrin

ello

-Rah

man

baro

stat

with

the

Nos

e-H

oove

rth

erm

o-st

at,b

utto

keep

itsi

mpl

ew

eon

lysh

owth

eP

arrin

ello

-Rah

man

mod

ifica

tion

here

:

d2r

i

dt2

=F

i

mi−

Mdr

i

dt,

(3.3

2)

M=

b−

1

[ bdb′

dt+

db dtb′] b

′−1.

(3.3

3)

The

(inve

rse)

mas

spa

ram

eter

mat

rixW−

1de

term

ines

the

stre

ngth

ofth

eco

uplin

g,an

dho

wth

ebo

xca

nbe

defo

rmed

.T

hebo

xre

stric

tion

(3.

1)w

illbe

fulfi

lled

auto

mat

ical

lyif

the

corr

espo

ndin

gel

emen

tsof

W−

1ar

eze

ro.

Sin

ceth

eco

uplin

gst

reng

thal

sode

pend

son

the

size

ofyo

urbo

x,w

epr

efer

toca

lcul

ate

itau

tom

atic

ally

inG

RO

MA

CS

.You

only

have

topr

ovid

eth

eap

prox

imat

eis

othe

rmal

com

pres

sibi

litie

sβan

dth

epr

essu

retim

eco

nsta

ntτ pin

the

inpu

tfile

(Lis

the

larg

est

box

mat

rixel

emen

t):

( W−

1) ij

=4π

ij

3τ2 pL.

(3.3

4)

Just

asfo

rth

eN

ose-H

oove

rth

erm

osta

t,yo

ush

ould

real

ize

that

the

Par

rinel

lo-R

ahm

antim

eco

n-st

ant

isn

ote

quiv

alen

tto

the

rela

xatio

ntim

eus

edin

the

Ber

ends

enpr

essu

reco

uplin

gal

gorit

hm.

Inm

ostc

ases

you

will

need

tous

ea

4–5

times

larg

ertim

eco

nsta

ntw

ithP

arrin

ello

-Rah

man

cou-

plin

g.If

your

pres

sure

isve

ryfa

rfr

omeq

uilib

rium

,the

Par

rinel

lo-R

ahm

anco

uplin

gm

ayre

sult

inve

ryla

rge

box

osci

llatio

nsth

atco

uld

even

cras

hyo

urru

n.In

that

case

you

wou

ldha

veto

incr

ease

the

time

cons

tant

,or

(bet

ter)

use

the

wea

kco

uplin

gsc

hem

eto

reac

hth

eta

rget

pres

sure

,and

then

switc

hto

Par

rinel

lo-R

ahm

anco

uplin

gon

ceth

esy

stem

isin

equi

libriu

m.

Sur

face

tens

ion

coup

ling

Whe

na

perio

dic

syst

emco

nsis

tsof

mor

eth

anon

eph

ase,

sepa

rate

dby

surf

aces

whi

char

epa

r-al

lelt

oth

exy

-pla

ne,

the

surf

ace

tens

ion

and

the

z-co

mpo

nent

ofth

epr

essu

reca

nbe

coup

led

toa

pres

sure

bath

.P

rese

ntly

,th

ison

lyw

orks

with

the

Ber

ends

enpr

essu

reco

uplin

gal

gorit

hmin

GR

OM

AC

S.T

heav

erag

esu

rfac

ete

nsio

nγ(t

)ca

nbe

calc

ulat

edfr

omth

edi

ffere

nce

betw

een

the

norm

alan

dth

ela

tera

lpre

ssur

e:

γ(t

)=

1 n

∫ L z 0

{ Pzz(z,t

)−P

xx(z,t

)+P

yy(z,t

)2

} dz(3

.35)

=L

z n

{ Pzz(t

)−P

xx(t

)+P

yy(t

)2

}(3

.36)

Page 46: CS USER AL Simulations - GROMACS

26

Ch

ap

ter

3.

Alg

orith

ms

where

Lz

isthe

heightofthebox

andnis

thenum

berofsurfaces.

The

pressurein

thez-direction

iscorrected

byscaling

theheightofthe

boxw

ithµ

z :

∆P

zz

=∆t

τp{P

0zz −

Pzz (t)}

(3.37)

µzz

=1

zz ∆P

zz

(3.38)

This

issim

ilarto

normalpressure

coupling,except

thatthe

power

ofone

thirdis

missing.

The

pressurecorrection

inthe

z-directionis

thenused

toget

thecorrect

convergencefor

thesurface

tensionto

thereference

valueγ0 .T

hecorrection

factorfor

thebox-length

inthe

x/y-directionis:

µx/y

=1

+∆t

2τpβ

x/y (

0

µzz L

z− {

Pzz (t)

+∆P

zz −

Pxx (t)

+P

yy (t)

2

})(3.39)

The

valueofβ

zz

ism

orecritical

thanw

ithnorm

alpressure

coupling.N

ormally

anincorrect

compressibility

willjust

scaleτp ,

butw

ithsurface

tensioncoupling

itaffects

theconvergence

ofthe

surfacetension.

Whenβz

zis

settozero

(constantboxheight),

∆P

zis

alsosetto

zero,which

isnecessary

forobtaining

thecorrectsurface

tension.

The

complete

updatealgorithm

The

complete

algorithmfor

theupdate

ofvelocities

andcoordinates

isgiven

inF

ig.3.7.

The

SH

AK

Ealgorithm

ofstep4

isexplained

below.

GR

OM

AC

Shas

aprovision

to”freeze”

(preventm

otionof)

selectedparticles,

which

must

bedefined

asa

’freezegroup’.

This

isim

plemented

usinga

free

zefa

ctorf

g ,w

hichis

avector,

anddiffers

foreachfre

ezeg

rou

p(seesec.3.3).

This

vectorcontains

onlyzero

(freeze)or

one(don’t

freeze).W

henw

etake

thisfreeze

factorand

theexternalacceleration

ah

intoaccountthe

updatealgorithm

forthe

velocitiesbecom

es:

v(t+∆t2)

=f

g ∗λ∗ [v(t−

∆t2)+

F(t)m

∆t+

ah ∆

t ](3.40)

where

gand

hare

groupindices

which

differper

atom.

3.4.7O

utputstep

The

important

outputof

theM

Drun

isthetra

jecto

ryfile

na

me

.trjw

hichcontains

particlecoordinates

and-optionally-

velocitiesat

regularintervals.

Since

thetrajectory

filesare

lengthy,one

shouldnotsave

everystep!

Toretain

allinformation

itsufficesto

write

afram

eevery

15steps,

sinceatleast30

stepsare

made

perperiod

ofthehighestfrequency

inthe

system,and

Shannon’s

sampling

theoremstates

thattw

osam

plesper

periodof

thehighest

frequencyin

aband-lim

itedsignal

containall

availableinform

ation.B

utthat

stillgives

verylong

files!S

o,if

thehighest

frequenciesare

notofinterest,10or

20sam

plesper

psm

aysuffice.

Be

aware

ofthedistortion

ofhigh-frequency

motions

bythestro

bo

scop

ice

ffect,calleda

liasin

g:higherfrequenciesare

mirrored

with

respecttothe

sampling

frequencyand

appearas

lower

frequencies.

E.3

2.

gn

me

ig2

03

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-tuenum

ps

Tim

eunit:p

s,fs

,ns

,us

,ms,s

,morh

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-typ

eenum

no

Com

putediffusion

coefficientinone

direction:n

o,x

,yorz

-late

ral

enumn

oC

alculatethe

lateraldiffusionin

aplane

perpendicularto:

no

,x,y

orz-n

gro

up

int1

Num

berofgroups

tocalculate

MS

Dfor

-mw

boolye

sM

assw

eightedM

SD

-tresta

rttim

e0

Tim

ebetw

eenrestarting

pointsin

trajectory(ps)

-be

gin

fittim

e0

Starttim

efor

fittingthe

MS

D(ps)

-en

dfit

time

-1E

ndtim

efor

fittingthe

MS

D(ps),-1

istillend

E.32

gnm

eig

gnm

eigcalculates

theeigenvectors/values

ofa(H

essian)matrix,w

hichcan

becalculated

with

nm

run

.T

heeigenvectors

arew

rittento

atrajectory

file(

-v).

The

structureis

written

firstw

itht=

0.T

heeigenvectors

arew

rittenas

frames

with

theeigenvector

number

astim

estamp.

The

eigenvectorscan

beanalyzed

with

ga

na

eig

.A

nensem

bleofstructures

canbe

generatedfrom

theeigenvectors

with

gn

me

ns.

Files

-fh

essia

n.m

txInput

Hessian

matrix

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-o

eig

en

val.xvg

Output

xvgr/xmgr

file-v

eig

en

vec.trr

Output

Fullprecision

trajectory:trr

trj

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-mbool

yes

Divide

elements

ofH

essianby

productof

sqrt(mass)

ofinvolved

atoms

priorto

diagonalization.T

hisshould

beused

for’N

ormalM

odes’analy-sis

-firstint

1F

irsteigenvectorto

write

away

-last

int1

00

Lasteigenvectorto

write

away

E.33

gnm

ens

gn

me

ns

generatesan

ensemble

aroundan

averagestructure

ina

subspacew

hichis

definedby

aset

ofnorm

almodes

(eigenvectors).T

heeigenvectors

areassum

edto

bem

ass-weighted.

The

positionalong

eacheigenvector

israndom

lytaken

froma

Gaussian

distributionw

ithvariance

kT/eigenvalue.

By

defaultthe

startingeigenvector

isset

to7,

sincethe

firstsix

normal

modes

arethe

translationaland

rotationaldegreesoffreedom

.

Files

-ve

ige

nve

c.trrInput

Fullprecision

trajectory:trr

trj-e

eig

en

val.xvg

Inputxvgr/xm

grfile

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb

Page 47: CS USER AL Simulations - GROMACS

20

2A

pp

en

dix

E.

Ma

nu

alP

age

s

E.3

0g

mor

ph

gm

orph

does

alin

ear

inte

rpol

atio

nof

conf

orm

atio

nsin

orde

rto

crea

tein

term

edia

tes.

Ofc

ours

eth

ese

are

com

plet

ely

unph

ysic

al,b

utth

atyo

um

aytr

yto

just

ifyyo

urse

lf.O

utpu

tis

inth

efo

rmof

age

neric

traj

ecto

ry.

The

num

ber

ofin

term

edia

tes

can

beco

ntro

lled

with

the

-nin

term

flag.

The

first

and

last

flag

corr

espo

ndto

the

way

ofin

terp

olat

ing:

0co

rres

pond

sto

inpu

tstr

uctu

re1

whi

le1

corr

espo

nds

toin

puts

truc

utre

2.If

you

spec

ifyfir

st<

0or

last>

1ex

trap

olat

ion

will

beon

the

path

from

inpu

tstr

uctu

rex1

tox2

.In

gene

ralt

heco

ordi

nate

sof

the

inte

rmed

iate

x(i)

outo

fNto

tali

nter

mid

ates

corr

espo

ndto

:

x(i)

=x1

+(fi

rst+

(i/(N

-1))

*(la

st-fi

rst)

)*(x

2-x1

)

Fin

ally

the

RM

SD

with

resp

ectt

obo

thin

puts

truc

ture

sca

nbe

com

pute

dif

expl

icitl

yse

lect

ed(-

orop

tion)

.In

that

case

anin

dex

file

may

bere

adto

sele

ctw

hatg

roup

RM

Sis

com

pute

dfr

om.

File

s -f1

con

f1.g

roIn

put

Gen

eric

stru

ctur

e:gr

og9

6pd

btp

rtp

btp

a-f

2co

nf2

.gro

Inpu

tG

ener

icst

ruct

ure:

gro

g96

pdb

tpr

tpb

tpa

-oin

term

.xtc

Out

put

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b-o

rrm

s-in

term

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-nin

term

int

11

Num

ber

ofin

term

edia

tes

-first

real

0C

orre

spon

dsto

first

gene

rate

dst

ruct

ure

(0is

inpu

tx0,

see

abov

e)-la

stre

al1

Cor

resp

onds

tola

stge

nera

ted

stru

ctur

e(1

isin

putx

1,se

eab

ove)

-fit

bool

yes

Do

ale

asts

quar

esfit

ofth

ese

cond

toth

efir

stst

ruct

ure

befo

rein

terp

olat

-in

g

E.3

1g

msd

gm

sdco

mpu

tes

the

mea

nsq

uare

disp

lace

men

t(M

SD

)of

atom

sfr

omth

eir

initi

alpo

sitio

ns.

Thi

spr

ovid

esan

easy

way

toco

mpu

teth

edi

ffusi

onco

nsta

ntus

ing

the

Ein

stei

nre

latio

n.T

hedi

ffusi

onco

nsta

ntis

calc

u-la

ted

byle

ast

squa

res

fittin

ga

stra

ight

line

thro

ugh

the

MS

Dfr

om-be

gin

fitto

-en

dfit

.A

ner

ror

estim

ate

give

n,w

hich

isth

edi

ffere

nce

ofth

edi

ffusi

onco

effic

ient

sob

tain

edfr

omfit

sov

erth

etw

oha

lfsof

the

fitin

terv

al.

Opt

ion

-mo

lpl

ots

the

MS

Dfo

rm

olec

ules

,th

isim

plie

s-mw

,i.e

.fo

rea

chin

idiv

idua

lmol

ecul

ean

diffu

-si

onco

nsta

ntis

com

pute

d.W

hen

usin

gan

inde

xfil

e,it

shou

ldco

ntai

nm

olec

ule

num

bers

inst

ead

ofat

omnu

mbe

rs.

Usi

ngth

isop

tion

one

also

gets

anac

cura

teer

ror

estim

ate

base

don

the

stat

istic

sbe

twee

nin

divi

d-ua

lmol

ecul

es.

Sin

ceon

eus

ually

isin

tere

sted

inse

lf-di

ffusi

onat

infin

itedi

lutio

nth

isis

prob

ably

the

mos

tus

eful

num

ber.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Str

uctu

re+

mas

s(db

):tp

rtp

btp

agr

og9

6pd

b-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-om

sd.x

vgO

utpu

txv

gr/x

mgr

file

-mo

ld

iffm

ol.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e

3.4

.M

ole

cula

rD

yna

mic

s2

7

TH

EU

PD

ATE

ALG

OR

ITH

M

Giv

en:

Pos

ition

srof

alla

tom

sat

timet

Velo

citie

svof

alla

tom

sat

timet−

∆t

2A

ccel

erat

ions

F/m

onal

lato

ms

attim

et.(F

orce

sar

eco

mpu

ted

disr

egar

ding

any

cons

trai

nts)

Tota

lkin

etic

ener

gyan

dvi

rial

⇓1.

Com

pute

the

scal

ing

fact

orsλan

acco

rdin

gto

eqns

.3.23

and

3.29

⇓2.

Upd

ate

and

scal

eve

loci

ties:v′=λ(v

+a∆t)

⇓3.

Com

pute

new

unco

nstr

aine

dco

ordi

nate

s:r′=

r+

v′ ∆t

⇓4.

App

lyco

nstr

aint

algo

rithm

toco

ordi

nate

s:co

nstr

ain(

r′→

r′′ ;

r)

⇓5.

Cor

rect

velo

citie

sfo

rco

nstr

aint

s:v=

(r′′−

r)/

∆t

⇓6.

Sca

leco

ordi

nate

san

dbo

x:r=µr′′ ;

b=µb

Fig

ure

3.7:

The

MD

upda

teal

gorit

hm

Page 48: CS USER AL Simulations - GROMACS

28

Ch

ap

ter

3.

Alg

orith

ms

3.5S

hellmolecular

dynamics

GR

OM

AC

Scan

simulate

polarizabilityusing

theshell

model

ofD

ickand

Overhauser

[23].

Insuch

models

ashellparticle

representingthe

electronicdegrees

offreedomis

attachedto

anucleus

bya

spring.T

hepotential

energyis

minim

izedw

ithrespect

tothe

shellposition

atevery

stepof

thesim

ulation(see

below).

Succesfullapplications

ofshellm

odelsin

GR

OM

AC

Shave

beenpublished

forN2

[24]andw

ater[25].

3.5.1O

ptimization

oftheshellpositions

The

forceFS

ona

shellparticleScan

bedecom

posedinto

two

components:

FS

=F

bond

+F

nb

(3.41)

whereF

bond

denotesthe

componentrepresenting

thepolarization

energy,usuallyrepresented

bya

harmonic

potentialandFnb

isthe

sumofC

oulomb

andVan

derW

aalsinteractions.

Ifwe

assume

thatFnb

isalm

ostconstantwe

cananalytically

derivethe

optimalposition

oftheshell,i.e.

where

FS

=0.

Ifwe

havethe

shellSconnected

toatom

Aw

ehave

Fbo

nd

=k

b (xS−

xA)

(3.42)

Inan

iterativesolver,

we

havepositionsxS (n)

where

nis

theiteration

count.W

enow

haveit

iterationn

:

Fnb

=F

S−k

b (xS (n)−

xA)

(3.43)

andthe

optimalposition

forthe

shellsxS (n+

1)thus

follows

from

FS−k

b (xS (n)−

xA)+k

b (xS (n

+1)−

xA)

=0

(3.44)

ifwe

write

∆x

S=

xS (n

+1)−

xS (n)

(3.45)

we

finallyobtain

∆x

S=

FS/k

b(3.46)

which

thenyields

thealgorithm

tocom

putethe

nexttrialinthe

optimization

ofshellpositions:

xS (n

+1)

=x

S (n)+

FS/k

b(3.47)

3.6C

onstraintalgorithms

Constraints

canbe

imposed

inG

RO

MA

CS

usingLIN

CS

(default)or

thetraditional

SH

AK

Em

ethod.

E.2

8.

gm

dm

at

20

1

E.28

gm

dmat

gm

dmatm

akesdistance

matrices

consistingofthe

smallestdistance

between

residuepairs.

With

-frames

thesedistance

matrices

canbe

storedas

afunction

oftime,to

beable

tosee

differencesin

tertiarystructure

asa

funcionof

time.

Ifyou

chooseyour

optionsunw

ise,this

may

generatea

largeoutput

file.D

efaultonly

anaveraged

matrix

overthe

whole

trajectoryis

output.A

lsoa

countofthenum

berofdifferentatom

iccontacts

between

residuesoverthe

whole

trajectorycan

bem

ade.T

heoutputcan

beprocessed

with

xpm2ps

tom

akea

PostS

cript(tm)

plot.F

iles-f

traj.xtc

InputG

enerictrajectory:

xtctrr

trjgrog96

pdb-s

top

ol.tp

rInput

Structure+

mass(db):

tprtpb

tpagro

g96pdb

-nin

de

x.nd

xInput,O

pt.Index

file-m

ea

nd

m.xp

mO

utputX

PixM

apcom

patiblem

atrixfile

-fram

es

dm

f.xpm

Output,O

pt.X

PixM

apcom

patiblem

atrixfile

-no

nu

m.xvg

Output,O

pt.xvgr/xm

grfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-treal

1.5

truncdistance

-nle

vels

int4

0D

iscretizedistance

in#

levels

E.29

gm

indist

gm

indistcom

putesthe

distancebetw

eenone

groupand

anum

berof

othergroups.

Both

them

inimum

distanceand

thenum

berofcontacts

within

agiven

distanceare

written

totw

oseparate

outputfiles.

With

option-p

ithe

minim

umdistance

ofa

groupto

itsperiodic

image

isplotted.

This

isuseful

forchecking

ifa

proteinhas

seenits

periodicim

ageduring

asim

ulation.O

nlyone

shiftin

eachdirection

isconsidered,giving

atotalof26

shifts.Italso

plotsthe

maxim

umdistance

within

thegroup

andthe

lengthsofthe

threebox

vectors.T

hisoption

isvery

slow.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

Input,Opt.

Structure+

mass(db):

tprtpb

tpagro

g96pdb

-nin

de

x.nd

xInput,O

pt.Index

file-o

dm

ind

ist.xvgO

utputxvgr/xm

grfile

-on

nu

mco

nt.xvg

Output

xvgr/xmgr

file-o

atm

-pa

ir.ou

tO

utputG

enericoutputfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-m

atrix

booln

oC

alculatehalfa

matrix

ofgroup-groupdistances

-dreal

0.6

Distance

forcontacts

-pi

booln

oC

alculatem

inimum

distancew

ithperiodic

images

Page 49: CS USER AL Simulations - GROMACS

20

0A

pp

en

dix

E.

Ma

nu

alP

age

s

8.A

vera

geC

alph

a-C

alph

adi

hedr

alan

gle

(file

phi-a

hx.x

vg).

9.A

vera

geP

hian

dP

sian

gles

(file

phip

si.x

vg).

10.E

llipt

icity

at22

2nm

acco

rdin

gtoH

irsta

nd

Bro

oks

File

s-s

top

ol.t

pr

Inpu

tG

ener

icru

nin

put:

tpr

tpb

tpa

-nin

de

x.n

dx

Inpu

tIn

dex

file

-ftr

aj.x

tcIn

put

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b-t

og

tra

j.g8

7O

utpu

t,O

pt.

Gro

mos

-87

AS

CII

traj

ecto

ryfo

rmat

-cz

zco

nf.g

roO

utpu

tG

ener

icst

ruct

ure:

gro

g96

pdb

-co

wa

ver.

gro

Out

put

Gen

eric

stru

ctur

e:gr

og9

6pd

b

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-r

0in

t1

The

first

resi

due

num

ber

inth

ese

quen

ce-q

bool

no

Che

ckat

ever

yst

epw

hich

part

ofth

ese

quen

ceis

helic

al-F

bool

yes

Togg

lefit

toa

perf

ecth

elix

-db

bool

no

Prin

tdeb

ugin

fo-e

vbo

oln

oW

rite

ane

w’tr

ajec

tory

’file

for

ED

-ah

xsta

rtin

t0

Firs

tres

idue

inhe

lix-a

hxe

nd

int

0La

stre

sidu

ein

helix

E.2

7g

lie

glie

com

pute

sa

free

ener

gyes

timat

eba

sed

onan

ener

gyan

alys

isfr

om.

One

need

san

ener

gyfil

ew

ithth

efo

llow

ing

com

pone

nts:

Cou

l(A

-B)

LJ-S

R(A

-B)

etc.

File

s-f

en

er.

ed

rIn

put

Gen

eric

ener

gy:

edr

ene

-olie

.xvg

Out

put

xvgr

/xm

grfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-E

ljre

al0

Lenn

ard-

Jone

sin

tera

ctio

nbe

twee

nlig

and

and

solv

ent

-Eq

qre

al0

Cou

lom

bin

tera

ctio

nbe

twee

nlig

and

and

solv

ent

-Clj

real

0.1

81

Fac

tor

inth

eLI

Eeq

uatio

nfo

rLe

nnar

d-Jo

nes

com

pone

ntof

ener

gy-C

qq

real

0F

acto

rin

the

LIE

equa

tion

for

Cou

lom

bco

mpo

nent

ofen

ergy

-lig

an

dst

ring

no

ne

Nam

eof

the

ligan

din

the

ener

gyfil

e

3.6

.C

on

stra

inta

lgo

rith

ms

29

3.6.

1S

HA

KE

The

SH

AK

E[2

6]al

gorit

hmch

ange

sa

seto

func

onst

rain

edco

ordi

nate

sr′

toa

seto

fcoo

rdin

ates

r′′

that

fulfi

lla

listo

fdis

tanc

eco

nstr

aint

s,us

ing

ase

tr

asre

fere

nce:

SH

AK

E(r

′→

r′′ ;

r)

Thi

sac

tion

isco

nsis

tent

with

solv

ing

ase

tofL

agra

nge

mul

tiplie

rsin

the

cons

trai

ned

equa

tions

ofm

otio

n.S

HA

KE

need

sato

lera

nce

TO

L;it

will

cont

inue

until

allc

onst

rain

tsar

esa

tisfie

dw

ithin

are

lativ

eto

lera

nceT

OL.

An

erro

rm

essa

geis

give

nif

SH

AK

Eca

nnot

rese

tthe

coor

dina

tes

beca

use

the

devi

atio

nis

too

larg

e,or

ifa

give

nnu

mbe

rof

itera

tions

issu

rpas

sed.

Ass

ume

the

equa

tions

ofm

otio

nm

ustf

ulfil

lK

holo

nom

icco

nstr

aint

s,ex

pres

sed

as

σk(r

1...r

N)

=0;

k=

1...K

(3.4

8)

(e.g

.(r

1−

r2)2−b2

=0)

.T

hen

the

forc

esar

ede

fined

as

−∂ ∂r

i

( V+

K ∑ k=

1

λkσ

k

)(3

.49)

whe

reλ

kar

eLa

gran

gem

ultip

liers

whi

chm

ust

beso

lved

tofu

lfill

the

cons

trai

nteq

uatio

ns.

The

seco

ndpa

rtof

this

sum

dete

rmin

esth

eco

nst

rain

tfo

rce

sGi,

defin

edby

Gi=−

K ∑ k=

1

λk∂σ

k

∂r

i(3

.50)

The

disp

lace

men

tdu

eto

the

cons

trai

ntfo

rces

inth

ele

apfr

ogor

Verle

tal

gorit

hmis

equa

lto

(Gi/m

i)(∆t)

2.

Sol

ving

the

Lagr

ange

mul

tiplie

rs(a

ndhe

nce

the

disp

lace

men

ts)

requ

ires

the

so-

lutio

nof

ase

tofc

oupl

edeq

uatio

nsof

the

seco

ndde

gree

.T

hese

are

solv

edite

rativ

ely

byS

HA

KE

.F

orth

esp

ecia

lcas

eof

rigid

wat

erm

olec

ules

,tha

tofte

nm

ake

upm

ore

than

80%

ofth

esi

mul

atio

nsy

stem

we

have

impl

emen

ted

the

SE

TT

LEal

gorit

hm[

27](

sec.

5.4)

.

3.6.

2LI

NC

S

The

LIN

CS

algo

rithm

LIN

CS

isan

algo

rithm

that

rese

tsbo

nds

toth

eirc

orre

ctle

ngth

saf

tera

nun

cons

trai

ned

upda

te[

28].

The

met

hod

isno

n-ite

rativ

e,as

ital

way

sus

estw

ost

eps.

Alth

ough

LIN

CS

isba

sed

onm

atric

es,

nom

atrix

-mat

rixm

ultip

licat

ions

are

need

ed.

The

met

hod

ism

ore

stab

lean

dfa

ster

than

SH

AK

E,

but

itca

non

lybe

used

with

bond

cons

trai

nts

and

isol

ated

angl

eco

nstr

aint

s,su

chas

the

prot

onan

gle

inO

H.B

ecau

seof

itsst

abili

tyLI

NC

Sis

espe

cial

lyus

eful

for

Bro

wni

andy

nam

ics.

LIN

CS

has

two

para

met

ers,

whi

char

eex

plai

ned

inth

esu

bsec

tion

para

met

ers.

The

LIN

CS

form

ulas

We

cons

ider

asy

stem

ofNpa

rtic

les,

with

posi

tions

give

nby

a3Nve

ctor

r(t

).F

orm

olec

ular

dyna

mic

sth

eeq

uatio

nsof

mot

ion

are

give

nby

New

ton’

sla

w

d2r

dt2

=M

−1F

(3.5

1)

Page 50: CS USER AL Simulations - GROMACS

30

Ch

ap

ter

3.

Alg

orith

ms

� �� �

� �� �

������������

� �� �

����

������������

� � �

�������

�������

�������

���������

���������

���������

���������

���������

���������

���������

���������

���������

���������

���������

���������

unconstrainedupdate

correction forrotational

lengthening

projecting outforces w

orkingalong the bonds

θ

d

ld

pd

Figure

3.8:T

hethree

positionupdates

neededfor

onetim

estep.

The

dashedline

isthe

oldbond

oflengthd,the

solidlines

arethe

newbonds.

l=d

cosθ

andp

=(2d

2−l 2)

12.

where

Fis

the3Nforce

vectorandM

isa3N

×3N

diagonalmatrix,containing

them

assesof

theparticles.

The

systemis

constrainedbyK

time-independentconstraintequations

gi (r)

=|r

i1 −r

i2 |−d

i =0

i=

1,...,K(3.52)

Ina

numerical

integrationschem

eLIN

CS

isapplied

afteran

unconstrainedupdate,

justlike

SH

AK

E.T

healgorithm

works

intw

osteps

(seefigure

Fig.

3.8).In

thefirststep

theprojections

ofthe

newbonds

onthe

oldbonds

areset

tozero.

Inthe

secondstep

acorrection

isapplied

forthe

lengtheningof

thebonds

dueto

rotation.T

henum

ericsfor

thefirst

stepand

thesecond

stepare

verysim

ilar.A

complete

derivationofthe

algorithmcan

befound

in[

28].O

nlya

shortdescriptionofthe

firststepis

givenhere.

Anew

notationis

introducedfor

thegradientm

atrixofthe

constraintequationsw

hichappears

onthe

righthandside

oftheequation

Bhi =

∂gh

∂ri

(3.53)

Notice

thatBis

aK×

3Nm

atrix,itcontainsthe

directionsofthe

constraints.T

hefollow

ingequa-

tionshow

show

thenew

constrainedcoordinates

rn+

1are

relatedto

theunconstrained

coordinatesr

unc

n+

1

rn+

1=

(I−

TnB

n )runc

n+

1+

Tnd

=

runc

n+

1 −M

−1B

n (BnM

−1B

Tn ) −1(B

nr

unc

n+

1 −d)

(3.54)

where

T=

M−

1BT(B

M−

1BT) −

1.T

hederivation

ofthis

equationfrom

eqns.3.51

and3.52

canbe

foundin

[28].

This

firststepdoes

notsettherealbond

lengthsto

theprescribed

lengths,buttheprojection

ofthenew

bondsonto

theold

directionsofthe

bonds.To

correctfortherotation

ofbondi,the

projectionofthe

bondon

theold

directionis

settopi = √

2d

2i −l 2i

(3.55)

where

liis

thebond

lengthafter

thefirstprojection.

The

correctedpositions

are

r∗n+

1=

(I−

TnB

n )rn+

1+

Tnp

(3.56)

E.2

6.

gh

elix

19

9

insertioninto

hydrogenbonds.

Ordering

isidenticalto

thatin-h

bn

indexfile.

-da

:w

riteoutthe

number

ofdonorsand

acceptorsanalyzed

foreach

timefram

e.T

hisis

especiallyusefull

when

using-she

ll.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-nin

de

x.nd

xInput,O

pt.Index

file-g

hb

on

d.lo

gO

utput,Opt.

Logfile

-sel

sele

ct.nd

xInput,O

pt.Index

file-n

um

hb

nu

m.xvg

Output

xvgr/xmgr

file-a

ch

ba

c.xvgO

utput,Opt.

xvgr/xmgr

file-d

isth

bd

ist.xvgO

utput,Opt.

xvgr/xmgr

file-a

ng

hb

an

g.xvg

Output,O

pt.xvgr/xm

grfile

-hx

hb

he

lix.xvgO

utput,Opt.

xvgr/xmgr

file-h

bn

hb

on

d.n

dx

Output,O

pt.Index

file-h

bm

hb

ma

p.xp

mO

utput,Opt.

XP

ixMap

compatible

matrix

file-d

ad

an

um

.xvgO

utput,Opt.

xvgr/xmgr

file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-ins

booln

oA

nalyzesolventinsertion

-areal

60

Cutoffangle

(degrees,Donor

-H

ydrogen-

Acceptor)

-rreal

0.2

5C

utoffradius(nm

,Hydrogen

-A

cceptor)-a

bin

real1

Binw

idthangle

distribution(degrees)

-rbin

real0

.00

5B

inwidth

distancedistribution

(nm)

-nita

ccbool

yes

Regard

nitrogenatom

sas

acceptors-sh

ell

real-1

when

>0,

onlycalculate

hydrogenbonds

within

#nm

shellaroundone

particle

E.26

ghelix

ghelix

computes

allkindof

helixproperties.

First,

thepeptide

ischecked

tofind

thelongest

helicalpart.T

hisis

determined

byH

ydrogenbonds

andP

hi/Psiangles.

That

bitis

fittedto

anidealhelix

aroundthe

Z-axis

andcentered

aroundthe

origin.T

henthe

following

propertiesare

computed:

1.Helix

radius(file

radius.xvg).T

hisis

merely

theR

MS

deviationin

two

dimensions

forallC

alphaatom

s.itis

calcedas

sqrt((SU

Mi(x 2(i)+

y2(i)))/N

),where

Nis

thenum

berofbackbone

atoms.

For

anidealhelix

theradius

is0.23

nm2.

Twist

(filetw

ist.xvg).T

heaverage

helicalangle

perresidue

iscalculated.

For

alphahelix

itis

100degrees,for

3-10helices

itwillbe

smaller,for

5-helicesitw

illbelarger.

3.R

iseper

residue(file

rise.xvg).T

hehelicalrise

perresidue

isplotted

asthe

differencein

Z-coordinate

between

Ca

atoms.

For

anidealhelix

thisis

0.15nm

4.Totalhelixlength

(filelen-ahx.xvg).

The

totallengthofthe

helixin

nm.

This

issim

plythe

averagerise

(seeabove)

times

thenum

berofhelicalresidues

(seebelow

).5.N

umber

ofhelicalresidues(file

n-ahx.xvg).T

hetitle

saysitall.

6.Helix

Dipole,backbone

only(file

dip-ahx.xvg).7.R

MS

deviationfrom

idealhelix,calculatedfor

theC

alphaatom

sonly

(filerm

s-ahx.xvg).

Page 51: CS USER AL Simulations - GROMACS

19

8A

pp

en

dix

E.

Ma

nu

alP

age

s

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-d

strin

gZ

Take

the

norm

alon

the

mem

bran

ein

dire

ctio

nX

,Yor

Z.

-sl

int

0C

alcu

late

orde

rpa

ram

eter

asfu

nctio

nof

boxl

engt

h,di

vidi

ngth

ebo

xin

#nr

slic

es.

•T

hepr

ogra

mas

sign

sw

hole

wat

erm

olec

ules

toa

slic

e,ba

sed

onth

efir

stat

omof

thre

ein

the

inde

xfil

egr

oup.

Itas

sum

esan

orde

rO

,H,H

.Nam

eis

noti

mpo

rtan

t,bu

tthe

orde

ris

.If

this

dem

and

isno

tm

et,a

ssig

ning

mol

ecul

esto

slic

esis

diffe

rent

.

E.2

5g

hbon

d

ghb

ond

com

pute

san

dan

alyz

eshy

drog

enbo

nds.

Hyd

roge

nbo

nds

are

dete

rmin

edba

sed

oncu

toffs

for

the

angl

eD

onor

-H

ydro

gen

-A

ccep

tor

(zer

ois

exte

nded

)an

dth

edi

stan

ceH

ydro

gen

-A

ccep

tor.

OH

and

NH

grou

psar

ere

gard

edas

dono

rs,O

isan

acce

ptor

alw

ays,

Nis

anac

cept

orby

defa

ult,

butt

his

can

besw

itche

dus

ing-

nita

cc.

Dum

my

hydr

ogen

atom

sar

eas

sum

edto

beco

nnec

ted

toth

efir

stpr

eced

ing

non-

hydr

ogen

atom

.

You

need

tosp

ecify

two

grou

psfo

ran

alys

is,w

hich

mus

tbe

eith

erid

entic

alor

non-

over

lapp

ing.

All

hydr

o-ge

nbo

nds

betw

een

the

two

grou

psar

ean

alyz

ed.

Ifyo

use

t-sh

ell,

you

will

beas

ked

for

anad

ditio

nali

ndex

grou

pw

hich

shou

ldco

ntai

nex

actly

one

atom

.In

this

case

,onl

yhy

drog

enbo

nds

betw

een

atom

sw

ithin

the

shel

ldis

tanc

efr

omth

eon

eat

omar

eco

nsid

ered

.

Itis

also

poss

ible

toan

alys

esp

ecifi

chy

drog

enbo

nds

with

-se

l.

Thi

sin

dex

file

mus

tco

ntai

na

grou

pof

atom

trip

lets

Don

orH

ydro

gen

Acc

epto

r,in

the

follo

win

gw

ay:

[se

lect

ed

]2

02

12

42

52

62

91

36

Not

eth

atth

etr

iple

tsne

edno

tbe

onse

para

telin

es.

Eac

hat

omtr

iple

tsp

ecifi

esa

hydr

ogen

bond

tobe

anal

yzed

,not

eal

soth

atno

chec

kis

mad

efo

rth

ety

pes

ofat

oms.

-in

stu

rns

onco

mpu

ting

solv

enti

nser

tion

into

hydr

ogen

bond

s.In

this

case

anad

ditio

nalg

roup

mus

tbe

sele

cted

,spe

cify

ing

the

solv

entm

olec

ules

.

Out

put:

-nu

m:

num

ber

ofhy

drog

enbo

nds

asa

func

tion

oftim

e.-a

c:

aver

age

over

alla

utoc

orre

latio

nsof

the

exis

tenc

efu

nctio

ns(e

ither

0or

1)of

allh

ydro

gen

bond

s.-d

ist

:di

stan

cedi

strib

utio

nof

allh

ydro

gen

bond

s.-a

ng

:an

gle

dist

ribut

ion

ofal

lhyd

roge

nbo

nds.

-hx

:th

enu

mbe

rof

n-n+

ihyd

roge

nbo

nds

asa

func

tion

oftim

ew

here

nan

dn+

ista

ndfo

rre

sidu

enu

mbe

rsan

dir

ange

sfr

om0

to6.

Thi

sin

clud

esth

en-

n+3,

n-n+

4an

dn-

n+5

hydr

ogen

bond

sas

soci

ated

with

helic

esin

prot

eins

.-h

bn

:al

lsel

ecte

dgr

oups

,don

ors,

hydr

ogen

san

dac

cept

ors

fors

elec

ted

grou

ps,a

llhy

drog

enbo

nded

atom

sfr

omal

lgro

ups

and

alls

olve

ntat

oms

invo

lved

inin

sert

ion.

-hb

m:

exis

tenc

em

atrix

for

allh

ydro

gen

bond

sov

eral

lfra

mes

,th

isal

soco

ntai

nsin

form

atio

non

solv

ent

3.7

.S

imu

late

dA

nn

ea

ling

31

Thi

sco

rrec

tion

for

rota

tiona

lef

fect

sis

actu

ally

anite

rativ

epr

oces

s,bu

tdu

ring

MD

only

one

itera

tion

isap

plie

d.T

here

lativ

eco

nstr

aint

devi

atio

naf

ter

this

proc

edur

ew

illbe

less

than

0.00

01fo

rev

ery

cons

trai

nt.

Inen

ergy

min

imiz

atio

nth

ism

ight

notb

eac

cura

teen

ough

,so

the

num

ber

ofite

ratio

nsis

equa

lto

the

orde

rof

the

expa

nsio

n(s

eebe

low

).

Hal

foft

heC

PU

time

goes

toin

vert

ing

the

cons

trai

ntco

uplin

gm

atrix

BnM

−1B

T n,w

hich

has

tobe

done

ever

ytim

est

ep.

Thi

sK×K

mat

rixha

s1/m

i 1+

1/m

i 2on

the

diag

onal

.T

heof

f-di

agon

alel

emen

tsar

eon

lyno

n-ze

row

hen

two

bond

sar

eco

nnec

ted,

then

the

elem

ent

isco

sφ/m

c,

whe

rem

cis

the

mas

sof

the

atom

conn

ectin

gth

etw

obo

nds

and

φis

the

angl

ebe

twee

nth

ebo

nds.

The

mat

rixT

isin

vert

edth

roug

ha

pow

erex

pans

ion.

AK×K

mat

rixS

isin

trod

uced

whi

chis

the

inve

rse

squa

rero

otof

the

diag

onal

ofB

nM

−1B

T n.

Thi

sm

atrix

isus

edto

conv

ertt

hedi

agon

alel

emen

tsof

the

coup

ling

mat

rixto

one

(BnM

−1B

T n)−

1=

SS−

1(B

nM

−1B

T n)−

1S−

1S

=S

(SB

nM

−1B

T nS

)−1S

=S

(I−

An)−

1S

(3.5

7)

The

mat

rixA

nis

sym

met

rican

dsp

arse

and

has

zero

son

the

diag

onal

.T

hus

asi

mpl

etr

ick

can

beus

edto

calc

ulat

eth

ein

vers

e

(I−

An)−

1=

I+

An

+A

2 n+

A3 n

+...

(3.5

8)

Thi

sin

vers

ion

met

hod

ison

lyva

lidif

the

abso

lute

valu

esof

allt

heei

genv

alue

sof

An

are

smal

ler

than

one.

Inm

olec

ules

with

only

bond

cons

trai

nts

the

conn

ectiv

ityis

solo

wth

atth

isw

illal

way

sbe

true

,eve

nif

ring

stru

ctur

esar

epr

esen

t.P

robl

ems

can

aris

ein

angl

e-co

nstr

aine

dm

olec

ules

.B

yco

nstr

aini

ngan

gles

with

addi

tiona

ldis

tanc

eco

nstr

aint

sm

ultip

lesm

allr

ing

stru

ctur

esar

ein

tro-

duce

d.T

his

give

sa

high

conn

ectiv

ity,l

eadi

ngto

larg

eei

genv

alue

s.T

here

fore

LIN

CS

shou

ldN

OT

beus

edw

ithco

uple

dan

gle-

cons

trai

nts.

The

LIN

CS

Par

amet

ers

The

accu

racy

ofLI

NC

Sde

pend

son

the

num

ber

ofm

atric

esus

edin

the

expa

nsio

neq

n.3.

58.

For

MD

calc

ulat

ions

afo

urth

orde

rexp

ansi

onis

enou

gh.

For

Bro

wni

andy

nam

ics

with

larg

etim

est

eps

anei

ghth

orde

rex

pans

ion

may

bene

cess

ary.

The

orde

ris

apa

ram

eter

inth

ein

putfi

lefo

rm

dru

n.

The

impl

emen

tatio

nof

LIN

CS

isdo

nein

such

aw

ayth

atth

eal

gorit

hmw

illne

ver

cras

h.E

ven

whe

nit

isim

poss

ible

toto

rese

tthe

cons

trai

nts

LIN

CS

will

gene

rate

aco

nfor

mat

ion

whi

chfu

lfills

the

cons

trai

nts

asw

ella

spo

ssib

le.

How

ever

,LI

NC

Sw

illge

nera

tea

war

ning

whe

nin

one

step

abo

ndro

tate

sov

erm

ore

than

apr

edefi

ned

angl

e.T

his

angl

eis

set

byth

eus

erin

the

inpu

tfil

efo

rm

dru

n.

3.7

Sim

ulat

edA

nnea

ling

The

wel

lkno

wn

sim

ulat

edan

neal

ing

(SA

)pr

otoc

olis

impl

emen

ted

ina

sim

ple

way

into

GR

O-

MA

CS

.A

mod

ifica

tion

ofth

ete

mpe

ratu

reco

uplin

gsc

hem

eis

used

asa

very

basi

cim

plem

enta

-tio

nof

the

SA

algo

rithm

.T

hem

etho

dw

orks

asfo

llow

s:th

ere

fere

nce

tem

pera

ture

for

coup

ling

T0

(eqn

.3.2

2)is

notc

onst

antb

utca

nbe

varie

dlin

early

:

T0(s

tep)

=T

0∗

(λ0+

∆λ∗

step

)(3

.59)

Page 52: CS USER AL Simulations - GROMACS

32

Ch

ap

ter

3.

Alg

orith

ms

ifλ

0=

1and∆

λis

0this

isthe

plainM

Dalgorithm

.N

otethat

forstandard

SA

∆λ

must

benegative.

WhenT

0 (step)<0

itissetto

0,asnegative

temperatures

donothave

aphysicalm

eaning.T

his“feature”

allows

foran

annealingstrategy

inw

hichat

firstthe

temperature

isscaled

down

linearlyuntil0

K,and

when

more

stepsare

takenthe

simulation

proceedsat0

K.S

incethe

weak

couplingschem

edoes

notcouple

instantaneously,the

actualtemperature

willalw

aysbe

slightlyhigher

than0

K.

3.8S

tochasticD

ynamics

Stochastic

orvelocity

Langevindynam

icsadds

afriction

anda

noiseterm

toN

ewton’s

equationsofm

otion:

mi d

2ri

dt 2=−m

i ξi d

ri

dt

+F

i (r)+◦ri

(3.60)

where

ξi

isthe

frictionconstant[1/ps]

and◦ri (t)

isa

noiseprocess

with〈 ◦r

i (t)◦rj (t

+s)〉

=2m

i ξi k

BTδ(s)δ

ij .W

hen1/ξ

iis

largecom

paredto

thetim

escales

presentin

thesystem

,one

couldsee

stochasticdynam

icsas

molecular

dynamics

with

stochastictem

perature-coupling.T

headvantage

compared

toM

Dw

ithB

erendsentem

perature-couplingis

thatin

caseof

SD

thegen-

eratedensem

bleis

known.

For

vacuumsim

ulationsthere

isthe

additionaladvantagethatthere

isno

accumulation

oferrors

forthe

overalltranslationalandrotationaldegrees

offreedom

.W

hen1/ξ

iis

smallcom

paredto

thetim

escales

presentinthe

system,the

dynamics

willbe

completely

differentfromM

D,butthe

sampling

isstillcorrect.

GR

OM

AC

Suses

acom

plicatedthird-order

leap-frogalgorithm

[29]

tointegrate

equation(3.60).

When

constraintsare

presentinthe

system,tw

oconstraintsteps

areperform

edper

time

step.T

hekinetic

energyis

computed

atthew

holetim

estep,this

isdone

byaveraging

thevelocities

atminus

andplus

ahalftim

estep,w

itha

correctionfor

thefriction:

v(t)=

12 (v (

t−∆t2 )

+v (

t+∆t2 ))(

e −ξ∆

t+√

2 (1−e −

ξ∆

t ))(3.61)

Exactcontinuation

ofastochastic

dynamics

simulation

isnotpossible,since

apartfromthe

coor-dinates

andthe

velocitiesone

randomterm

oftheprevious

stepin

required,however,the

errorwill

bevery

small.

3.9B

rownian

Dynam

ics

Inthe

limitofhigh

frictionstochastic

dynamics

reducesto

Brow

niandynam

ics,also

calledposi-

tionLangevin

dynamics.

This

appliesto

over-damped

systems,

i.e.system

sin

which

theinertia

effectsare

negligible.T

heequation

is:

dri

dt=

1γi F

i (r)+◦ri

(3.62)

where

γi

isthe

frictioncoefficient[am

u/ps]and◦ri (t)

isa

noiseprocess

with〈 ◦r

i (t)◦rj (t+

s)〉=

2δ(s)δ

ij kBT/γ

i .In

GR

OM

AC

Sthe

equationsare

integratedw

itha

simple,explicitschem

e:

ri (t+

∆t)

=r

i (t)+

∆t

γiF

i (r(t))+ √

2kBT

∆t

γi

rGi

(3.63)

E.2

3.

gg

yrate

19

7

-nm

ol

int1

Num

berof

molecules

inyour

sample:

theenergies

aredivided

bythis

number

-nd

fint

3N

umber

ofdegrees

offreedom

perm

olecule.N

ecessaryfor

calculatingthe

heatcapacity-flu

cbool

no

Calculate

autocorrelationofenergy

fluctuationsrather

thanenergy

itself-a

cflen

int-1

Lengthofthe

AC

F,defaultishalfthe

number

offrames

-no

rma

lizebool

yes

Norm

alizeA

CF

-Penum

0O

rderofLegendre

polynomialfor

AC

F(0

indicatesnone):

0,1

,2or3

-fitfnenum

no

ne

Fitfunction:n

on

e,e

xp,a

exp

,exp

exp

orvac

-ncskip

int0

Skip

Npoints

inthe

outputfileofcorrelation

functions-b

eg

infit

real0

Tim

ew

hereto

beginthe

exponentialfitofthecorrelation

function-e

nd

fitreal

-1T

ime

where

toend

theexponentialfitofthe

correlationfunction,-1

istill

theend

E.23

ggyrate

ggyrate

computes

theradius

ofgyrationofa

groupofatom

sand

theradiiofgyration

aboutthex,y

andz

axes,asa

functionoftim

e.T

heatom

sare

explicitlym

assw

eighted.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-o

gyra

te.xvg

Output

xvgr/xmgr

file-n

ind

ex.n

dx

Input,Opt.

Indexfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-q

booln

oU

seabsolute

valueof

thecharge

ofan

atomas

weighting

factorinstead

ofmass

-pbool

no

Calculate

theradiiofgyration

abouttheprincipalaxes.

E.24

gh2order

Com

putethe

orientationofw

aterm

oleculesw

ithrespectto

thenorm

alofthebox.

The

programdeterm

inesthe

averagecosine

ofthe

anglebetw

eende

dipolem

oment

ofw

aterand

anaxis

ofthe

box.T

hebox

isdivided

inslices

andthe

averageorientation

perslice

isprinted.

Each

water

molecule

isassigned

toa

slice,per

time

frame,based

onthe

positionofthe

oxygen.W

hen-nm

isused

theangle

between

thew

aterdipole

andthe

axisfrom

thecenter

ofmass

tothe

oxygenis

calculatedinstead

oftheangle

between

thedipole

anda

boxaxis.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-nin

de

x.nd

xInput

Indexfile

-nm

ind

ex.n

dx

Input,Opt.

Indexfile

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-oo

rde

r.xvgO

utputxvgr/xm

grfile

Page 53: CS USER AL Simulations - GROMACS

19

6A

pp

en

dix

E.

Ma

nu

alP

age

s

-lj

bool

yes

calc

ulat

eLe

nnar

d-Jo

nes

SR

ener

gies

-lj1

4bo

oln

oca

lcul

ate

Lenn

ard-

Jone

s1-

4en

ergi

es-b

ha

mbo

oln

oca

lcul

ate

Buc

king

ham

ener

gies

-fre

ebo

olye

sca

lcul

ate

free

ener

gy-t

em

pre

al3

00

refe

renc

ete

mpe

ratu

refo

rfr

eeen

ergy

calc

ulat

ion

E.2

2g

ener

gy

gen

ergy

extr

acts

ener

gyco

mpo

nent

sor

dist

ance

rest

rain

tdat

afr

oman

ener

gyfil

e.T

heus

eris

prom

pted

toin

tera

ctiv

ely

sele

ctth

een

ergy

term

ssh

ew

ants

.

Whe

nth

e-vi

ol

optio

nis

set,

the

time

aver

aged

viol

atio

nsar

epl

otte

dan

dth

eru

nnin

gtim

e-av

erag

edan

din

stan

tane

ous

sum

ofvi

olat

ions

are

reca

lcul

ated

.A

dditi

onal

lyru

nnin

gtim

e-av

erag

edan

din

stan

tane

ous

dist

ance

sbe

twee

nse

lect

edpa

irsca

nbe

plot

ted

with

the

-pa

irs

optio

n.

Ave

rage

and

RM

SD

are

calc

ulat

edw

ithfu

llpr

ecis

ion

from

the

sim

ulat

ion

(see

prin

ted

man

ual).

Drif

tis

calc

ulat

edby

perf

orm

ing

aLS

Qfit

ofth

eda

tato

ast

raig

htlin

e.To

tald

rifti

sdr

iftm

ultip

lied

byto

talt

ime.

With

-fe

ea

free

ener

gyes

timat

eis

calc

ulat

edus

ing

the

form

ula:

G=

-ln<

e(E

/kT

)>

*kT

,w

here

kis

Bol

tzm

ann’

sco

nsta

nt,T

isse

tby-fe

tem

pan

dth

eav

erag

eis

over

the

ense

mbl

e(o

rtim

ein

atr

ajec

tory

).N

ote

that

this

isin

prin

cipl

eon

lyco

rrec

twhe

nav

erag

ing

over

the

who

le(B

oltz

man

n)en

sem

ble

and

usin

gth

epo

tent

iale

nerg

y.T

his

also

allo

ws

for

anen

trop

yes

timat

eus

ing

G=

H-

TS

,w

here

His

the

enth

alpy

(H=

U+

pV

)an

dS

entr

opy.

Whe

na

seco

nden

ergy

file

issp

ecifi

ed(

-f2

),a

free

ener

gydi

ffere

nce

isca

lcul

ated

dF=

-kT

ln<

e-(

EB

-E

A)/

kT>

A,

whe

reE

Aan

dE

Bar

eth

een

ergi

esfr

omth

efir

stan

dse

cond

ener

gyfil

es,

and

the

aver

age

isov

erth

een

sem

ble

A.NO

TE

that

the

ener

gies

mus

tbot

hbe

calc

ulat

edfr

omth

esa

me

traj

ecto

ry.

File

s-f

en

er.

ed

rIn

put

Gen

eric

ener

gy:

edr

ene

-f2

en

er.

ed

rIn

put,

Opt

.G

ener

icen

ergy

:ed

ren

e-s

top

ol.t

pr

Inpu

t,O

pt.

Gen

eric

run

inpu

t:tp

rtp

btp

a-o

en

erg

y.xv

gO

utpu

txv

gr/x

mgr

file

-vio

lvi

ola

ver.

xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-pa

irs

pa

irs.

xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-co

rre

ne

corr

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-vis

visc

o.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-r

avg

run

avg

df.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-fe

ebo

oln

oD

oa

free

ener

gyes

timat

e-f

ete

mp

real

30

0R

efer

ence

tem

pera

ture

for

free

ener

gyca

lcul

atio

n-z

ero

real

0S

ubtr

acta

zero

-poi

nten

ergy

-su

mbo

oln

oS

umth

een

ergy

term

sse

lect

edra

ther

than

disp

lay

them

all

-dp

bool

no

Prin

tene

rgie

sin

high

prec

isio

n-m

uto

tbo

oln

oC

ompu

teth

eto

tald

ipol

em

omen

tfro

mth

eco

mpo

nent

s-s

kip

int

0S

kip

num

ber

offr

ames

betw

een

data

poin

ts-a

ver

bool

no

Prin

tals

oth

eX

1,ta

ndsi

gma1

,t,on

lyif

only

1en

ergy

isre

ques

ted

3.1

0.

En

erg

yM

inim

iza

tion

33

whe

rer

G iis

Gau

ssia

ndi

strib

uted

nois

ew

ithµ=

0,σ

=1.

The

fric

tion

coef

ficie

ntsγ

ica

nbe

chos

enth

esa

me

for

allp

artic

les

orasγ i

=m

i/ξ i

,whe

reth

efr

ictio

nco

nsta

ntsξ i

can

bedi

ffere

ntfo

rdi

ffere

ntgr

oups

ofat

oms.

Bec

ause

the

syst

emis

assu

med

tobe

over

dam

ped,

larg

etim

e-st

eps

can

beus

ed.

LIN

CS

shou

ldbe

used

for

the

cons

trai

nts

sinc

eS

HA

KE

will

notc

onve

rge

for

larg

eat

omic

disp

lace

men

ts.

BD

isan

optio

nof

the

md

run

prog

ram

.

3.10

Ene

rgy

Min

imiz

atio

n

Ene

rgy

min

imiz

atio

nin

GR

OM

AC

Sca

nbe

done

usin

ga

stee

pest

desc

ent

orco

njug

ate

grad

ient

met

hod.

EM

isju

stan

optio

nof

them

dru

npr

ogra

m.

3.10

.1S

teep

estD

esce

nt

Alth

ough

stee

pest

desc

ent

isce

rtai

nly

not

the

mos

tef

ficie

ntal

gorit

hmfo

rse

arch

ing,

itis

robu

stan

dea

syto

impl

emen

t.

We

defin

eth

eve

ctorr

asth

eve

ctor

ofal

l3Nco

ordi

nate

s.In

itial

lya

max

imum

disp

lace

men

th

0

(e.g

.0.0

1nm

)m

ustb

egi

ven.

Firs

tthe

forc

esF

and

pote

ntia

lene

rgy

are

calc

ulat

ed.

New

posi

tions

are

calc

ulat

edby

rn+

1=

rn

+F

n

max

(|Fn|)h

n(3

.64)

whe

reh

nis

the

max

imum

disp

lace

men

tandF

nis

the

forc

e,or

the

nega

tive

grad

ient

ofth

epo

ten-

tialV

.T

heno

tatio

nmax

(|Fn|)

mea

nsth

ela

rges

toft

heab

solu

teva

lues

ofth

efo

rce

com

pone

nts.

The

forc

esan

den

ergy

are

agai

nco

mpu

ted

for

the

new

posi

tions

If(V

n+

1<V

n)

the

new

posi

tions

are

acce

pted

and

hn+

1=

1.2h

n.

If(V

n+

1≥V

n)

the

new

posi

tions

are

reje

cted

and

hn

=0.

2hn.

The

algo

rithm

stop

sw

hen

eith

era

user

spec

ified

num

ber

offo

rce

eval

uatio

nsha

sbe

enpe

rfor

med

(e.g

.10

0),

orw

hen

the

max

imum

ofth

eab

solu

teva

lues

ofth

efo

rce

(gra

dien

t)co

mpo

nent

sis

smal

ler

than

asp

ecifi

edva

lueε.

Sin

cefo

rce

trun

catio

npr

oduc

esso

me

nois

ein

the

ener

gyev

alua

-tio

n,th

est

oppi

ngcr

iterio

nsh

ould

notb

em

ade

too

tight

toav

oid

endl

ess

itera

tions

.A

reas

onab

leva

lue

forε

can

bees

timat

edfr

omth

ero

otm

ean

squa

refo

rce

fa

harm

onic

osci

llato

rwou

ldex

hibi

tat

ate

mpe

ratu

reTT

his

valu

eis

f=

2πν√

2mkT

(3.6

5)

whe

reν

isth

eos

cilla

tor

freq

uenc

y,mth

e(r

educ

ed)

mas

s,an

dkB

oltz

man

n’s

cons

tant

.F

ora

wea

kos

cilla

tor

with

aw

ave

num

ber

of10

0cm−

1an

da

mas

sof

10at

omic

units

,ata

tem

pera

ture

of1

K,f

=7.

7kJ

mol−

1nm

−1.

Ava

lue

forε

betw

een

1an

d10

isac

cept

able

.

3.10

.2C

onju

gate

Gra

dien

t

Con

juga

tegr

adie

ntis

slow

erth

anst

eepe

stde

scen

tin

the

early

stag

esof

the

min

imiz

atio

n,bu

tbe

com

esm

ore

effic

ient

clos

erto

the

ener

gym

inim

um.

The

para

met

ers

and

stop

crite

rion

are

the

sam

eas

fors

teep

estd

esce

nt.

Con

juga

tegr

adie

ntca

nno

tbe

used

with

cons

trai

nts

orfr

eeze

grou

ps.

Page 54: CS USER AL Simulations - GROMACS

34

Ch

ap

ter

3.

Alg

orith

ms

3.11N

ormalM

odeA

nalysis

Norm

almode

analysis[30,31,32]can

beperform

edusing

GR

OM

AC

S,by

diagonalizationofthe

mass-w

eightedH

essian:M

−1/2H

M−

1/2Q

2Q(3.66)

where

Mcontains

theatom

icm

asses,Q

containseigenvectors,andωcontains

thecorresponding

eigenvalues(frequencies).

First,the

Hessian

matrix,w

hichis

a3N×

3Nm

atrixw

hereNis

thenum

berofatom

s,hasto

becalculated:

Hij

=∂

2V

∂x

i ∂x

j(3.67)

where

xi and

xj

denotethe

atomic

x,yor

zcoordinates.

Inpractice,these

equationshave

notbeendeveloped

analytically,buttheforce

isused

Fi

=∂V

∂x

i(3.68)

fromw

hichthe

Hessian

iscom

putednum

erically.Itshould

benoted

thatforausualN

ormalM

odecalculation,itis

necessaryto

completely

minim

izethe

energypriorto

computation

oftheH

essian.T

hisshould

bedone

with

conjugategradient

indouble

precision.A

number

ofG

RO

MA

CS

pro-gram

sare

involvedin

thesecalculations.

Firstn

mru

n,w

hichcom

putesthe

Hessian,and

secondlyg

nm

eig

which

doesthe

diagonalizationand

sortingofnorm

almodes

accordingto

frequencies.B

oththese

programs

shouldbe

runin

doubleprecision.

An

overviewofnorm

almode

analysisand

therelated

principalcomponentanalysis

(seesec.8.9)

canbe

foundin

[33].

3.12F

reeenergy

calculations

Free

energycalculations

canbe

performed

inG

RO

MA

CS

usingslow

-growth

methods.

An

exam-

pleproblem

might

be:calculate

thedifference

infree

energyof

bindingof

aninhibitor

Ito

anenzym

eEand

toa

mutated

enzymeE’.Itis

notfeasiblew

ithcom

putersim

ulationsto

performa

dockingcalculation

forsuch

alarge

complex,or

evenreleasing

theinhibitor

fromthe

enzyme

ina

reasonableam

ountof

computer

time

with

reasonableaccuracy.

How

ever,if

we

considerthe

freeenergy

cyclein

(Fig.3.9A

)w

ecan

write

∆G

1 −∆G

2=

∆G

3 −∆G

4(3.69)

Ifwe

areinterested

inthe

left-handterm

we

canequally

wellcom

putethe

right-handterm

.

Ifw

ew

antto

compute

thedifference

infree

energyof

bindingof

two

inhibitorsI

andI’

toan

enzymeE

(Fig.3.9B

)w

ecan

againuse

eqn.3.69

tocom

putethe

desiredproperty.

Free

energydifferences

between

two

molecular

speciescan

becalculated

inG

RO

MA

CS

usingthe

“slow-grow

th”m

ethod.In

fact,suchfree

energydifferences

between

differentmolecular

speciesare

physicallym

eaningless,but

theycan

beused

toobtain

meaningful

quantitiesem

ployinga

thermodynam

iccycle.

The

method

requiresa

simulation

duringw

hichthe

Ham

iltonianof

thesystem

changesslow

lyfrom

thatdescribing

onesystem

(A)

tothat

describingthe

othersystem

E.2

1.

ge

ne

ma

t1

95

Files

-fd

ynd

om

.pd

bInput

Protein

databank

file-o

rota

ted

.xtcO

utputG

enerictrajectory:

xtctrr

trjgrog96

pdb-n

do

ma

ins.n

dx

InputIndex

file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int0

Setthe

nicelevel-firsta

ng

lereal

0A

ngleofrotation

aboutrotationvector

-lasta

ng

lereal

0A

ngleofrotation

aboutrotationvector

-nfra

me

int1

1N

umber

ofstepson

thepathw

ay-m

axa

ng

lereal

0D

ymD

omdterm

inedangle

ofrotationaboutrotation

vector-tra

ns

real0

Translation

(Aangstroem

)along

rotationvector

(seeD

ynDom

infofile)

-he

ad

vector0

00

Firstatom

ofthearrow

vector-ta

ilvector

00

0Lastatom

ofthearrow

vector

E.21

genem

at

genem

atextractsan

energym

atrixfrom

anenergy

file.W

ith-groups

afile

mustbe

suppliedw

ithon

eachline

agroup

tobe

used.F

orthese

groupsa

matrices

ofinteractionenergies

willbe

calculated.A

lsothe

totalinteraction

energyenergy

pergroup

iscalculated.

An

approximation

ofthe

freeenergy

iscalculated

using:E

(free)=

E0

+kT

log(<

exp((E-E

0)/kT)>

),w

here’<>

’standsfor

time-average.

Afile

with

referencefree

energiescan

besupplied

tocalculate

thefree

energydifference

with

some

referencestate.

Group

names

(e.g.residue

names

inthe

referencefile

shouldcorrespond

tothe

groupnam

esas

usedin

the-groups

file,but

aappended

number

(e.g.residue

number)in

the-groupsw

illbeignored

inthe

comparison.

Files

-fe

ne

r.ed

rInput,O

pt.G

enericenergy:

edrene

-gro

up

sg

rou

ps.d

at

InputG

enericdata

file-e

ref

ere

f.da

tInput,O

pt.G

enericdata

file-e

ma

te

ma

t.xpm

Output

XP

ixMap

compatible

matrix

file-e

tot

en

erg

y.xvgO

utputxvgr/xm

grfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-su

mbool

no

Sum

theenergy

terms

selectedrather

thandisplay

themall

-skipint

0S

kipnum

beroffram

esbetw

eendata

points-m

ea

nbool

yes

with

-groupscalculates

matrix

ofm

eanenergies

instead

ofm

atrixfor

eachtim

estep-n

leve

lsint

20

number

oflevelsfor

matrix

colors-m

ax

real1

e+

20

max

valuefor

energies-m

inreal

-1e

+2

0m

invalue

forenergies

-cou

lbool

yes

calculateC

oulomb

SR

energies-co

ulr

booln

ocalculate

Coulom

bLR

energies-co

ul1

4bool

no

calculateC

oulomb

1-4energies

Page 55: CS USER AL Simulations - GROMACS

19

4A

pp

en

dix

E.

Ma

nu

alP

age

s

-dm

drm

ax.

xvg

Out

put

xvgr

/xm

grfil

e-d

rre

str.

xvg

Out

put

xvgr

/xm

grfil

e-l

dis

res.

log

Out

put

Log

file

-nvi

ol.n

dx

Inpu

t,O

pt.

Inde

xfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-n

top

int

6N

umbe

rof

larg

evi

olat

ions

that

are

stor

edin

the

log

file

ever

yst

ep

E.1

9g

dist

gdi

stca

nca

lcul

ate

the

dist

ance

betw

een

the

cent

ers

ofm

ass

oftw

ogr

oups

ofat

oms

asa

func

tion

oftim

e.T

heto

tald

ista

nce

and

itsx,

yan

dz

com

pone

nts

are

plot

ted.

Or

whe

n-d

ist

isse

t,pr

inta

llth

eat

oms

ingr

oup

2th

atar

ecl

oser

than

ace

rtai

ndi

stan

ceto

the

cent

erof

mas

sof

grou

p1.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-od

ist.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-dis

tre

al0

Prin

tall

atom

sin

grou

p2

clos

erth

andi

stto

the

cent

erof

mas

sof

grou

p1

E.2

0g

dynd

om

gdy

ndom

read

sa

pdb

file

outp

utfr

omD

ynD

omht

tp://

md.

chem

.rug

.nl/

stev

e/D

ynD

om/d

yndo

m.h

ome.

htm

lIt

read

sth

eco

ordi

nate

s,an

dth

eco

ordi

nate

sof

the

rota

tion

axis

furt

herm

ore

itre

ads

anin

dex

file

cont

aini

ngth

edo

mai

ns.

Fur

ther

mor

eit

take

sth

efir

stan

dla

stat

omof

the

arro

wfil

eas

com

man

dlin

ear

gum

ents

(hea

dan

dta

il)an

dfin

ally

itta

kes

the

tran

slat

ion

vect

or(g

iven

inD

ynD

omin

fofil

e)an

dth

ean

gle

ofro

tatio

n(a

lso

asco

mm

and

line

argu

men

ts).

Ifth

ean

gle

dete

rmin

edby

Dyn

Dom

isgi

ven,

one

shou

ldbe

able

tore

cove

rth

ese

cond

stru

ctur

eus

edfo

rge

nera

ting

the

Dyn

Dom

outp

ut.

Bec

ause

oflim

ited

num

eric

alac

cura

cyth

issh

ould

beve

rified

byco

mpu

ting

anal

l-ato

mR

MS

D(u

singg

con

frm

s)

rath

erth

anby

file

com

paris

on(u

sing

diff)

.

The

purp

ose

ofth

ispr

ogra

mis

toin

terp

olat

ean

dex

trap

olat

eth

ero

tatio

nas

foun

dby

Dyn

Dom

.A

sa

resu

ltun

phys

ical

stru

ctur

esw

ithlo

ngor

shor

tbon

ds,o

rov

erla

ppin

gat

oms

may

bepr

oduc

ed.

Vis

uali

nspe

ctio

n,an

den

ergy

min

imiz

atio

nm

aybe

nece

ssar

yto

valid

ate

the

stru

ctur

e.

3.1

2.

Fre

ee

ne

rgy

calc

ula

tion

s3

5

I E’

EI EE

G1

∆∆G

2

∆G4

∆G3

A

G1

∆∆G

2

∆G3

II’

EI

EI’

∆G4

B

Fig

ure

3.9:

Fre

een

ergy

cycl

es.A:

toca

lcul

ate∆G

12

orth

efr

eeen

ergy

diffe

renc

ebe

twee

nth

ebi

ndin

gof

inhi

bito

rIto

enzy

mes

Ere

spec

tivel

yE’.

B:t

oca

lcul

ate∆G

12

whi

chis

the

free

ener

gydi

ffere

nce

for

bind

ing

ofin

hibi

torsI

resp

ectiv

elyI

’to

enzy

meE

.

(B).

The

chan

gem

ustb

eso

slow

that

the

syst

emre

mai

nsin

equi

libriu

mdu

ring

the

proc

ess;

ifth

atre

quire

men

tis

fulfi

lled,

the

chan

geis

reve

rsib

lean

da

slow

-gro

wth

sim

ulat

ion

from

Bto

Aw

illyi

eld

the

sam

ere

sults

(but

with

adi

ffere

ntsi

gn)

asa

slow

-gro

wth

sim

ulat

ion

from

Ato

B.T

his

isa

usef

ulch

eck,

butt

heus

ersh

ould

beaw

are

ofth

eda

nger

that

equa

lity

offo

rwar

dan

dba

ckw

ard

grow

thre

sults

does

notg

uara

ntee

corr

ectn

ess

ofth

ere

sults

.

The

requ

ired

mod

ifica

tion

ofth

eH

amilt

onia

nHis

real

ized

bym

akin

gHa

func

tion

ofac

ou

plin

gp

ara

me

terλ

:H=H

(p,q

;λ)

insu

cha

way

thatλ

=0

desc

ribes

syst

emA

andλ=

1de

scrib

essy

stem

B:

H(p,q

;0)

=H

A(p,q

);H

(p,q

;1)

=H

B(p,q

).(3

.70)

InG

RO

MA

CS

,th

efu

nctio

nal

form

ofth

eλ-d

epen

denc

eis

diffe

rent

for

the

vario

usfo

rce-

field

cont

ribut

ions

and

isde

scrib

edin

sect

ion

sec.

4.3.

The

Hel

mho

ltzfr

eeen

ergyA

isre

late

dto

the

part

ition

func

tionQ

ofanN,V,T

ense

mbl

e,w

hich

isas

sum

edto

beth

eeq

uilib

rium

ense

mbl

ege

nera

ted

bya

MD

sim

ulat

ion

atco

nsta

ntvo

lum

ean

dte

mpe

ratu

re.

The

gene

rally

mor

eus

eful

Gib

bsfr

eeen

ergy

Gis

rela

ted

toth

epa

rtiti

onfu

nctio

n∆

ofanN,p,T

ense

mbl

e,w

hich

isas

sum

edto

beth

eeq

uilib

rium

ense

mbl

ege

nera

ted

bya

MD

sim

ulat

ion

atco

nsta

ntpr

essu

rean

dte

mpe

ratu

re:

A(λ

)=

−k

BT

lnQ

(3.7

1)

Q=

c

∫∫ex

p[−βH

(p,q

;λ)]dpdq

(3.7

2)

G(λ

)=

−k

BT

ln∆

(3.7

3)

∆=

c

∫∫∫ex

p[−βH

(p,q

;λ)−βpV

]dpdqdV

(3.7

4)

G=

A+pV,

(3.7

5)

whe

reβ

=1/

(kBT

)an

dc

=(N

!h3N

)−1.

The

sein

tegr

als

over

phas

esp

ace

cann

otbe

eval

uate

dfr

oma

sim

ulat

ion,

but

itis

poss

ible

toev

alua

teth

ede

rivat

ive

toth

epa

ram

eter

λas

anen

sem

ble

Page 56: CS USER AL Simulations - GROMACS

36

Ch

ap

ter

3.

Alg

orith

ms

average:dAdλ

= ∫∫(∂H/∂λ)exp[−

βH

(p,q;λ)]d

pdq

∫∫exp[−

βH

(p,q;λ)]d

pdq

= ⟨∂H∂λ ⟩

NV

T;λ,

(3.76)

with

asim

ilarrelation

fordG/dλ

inthe

N,p,T

ensemble.

The

differencein

freeenergy

between

Aand

Bcan

befound

byintegrating

thederivative

overλ

:

AB(V

,T)−

AA(V,T

)= ∫

1

0 ⟨∂H∂λ ⟩

NV

T;λdλ

(3.77)

GB(p

,T)−

GA(p,T

)= ∫

1

0 ⟨∂H∂λ ⟩

NpT

;λdλ.

(3.78)

Ifone

wishes

toevaluateG

B(p,T

)−G

A(p,T

),the

naturalchoiceis

aconstant-pressure

simu-

lation.H

owever,

thisquantity

canalso

beobtained

froma

slow-grow

thsim

ulationat

constantvolum

e,startingw

ithsystem

Aatpressure

pand

volumeV

andending

with

systemB

atpressurep

B,by

applyingthe

following

smallcorrection:

GB(p)−

GA(p)

=A

B(V)−

AA(V

)− ∫pB

p[V

B(p′)−

V]dp′

(3.79)

Here

we

omitted

theconstant

Tfrom

thenotation.

This

correctionis

roughlyequalto

−12 (p

B−

p)∆V

=(∆V

)2/(2κ

V),w

here∆V

isthe

volume

changeatpand

κis

theisotherm

alcompress-

ibility.T

hisis

usuallynegligible.

For

example,the

growth

ofaw

aterm

oleculefrom

nothingin

abath

of1000w

aterm

oleculesatconstantvolum

ew

ouldproduce

anadditionalpressure

of22bar

anda

correctionto

theH

elmholtz

freeenergy

of-20J/m

ol.

Incartesian

coordinates,thekinetic

energyterm

inthe

Ham

iltoniandepends

onlyon

them

omenta,

andcan

beseparately

integratedand

infact

removed

fromthe

equations.W

henm

assesdo

notchange,there

isno

contributionfrom

thekinetic

energyatall;otherw

isethe

integratedcontribution

tothe

freeenergy

is−32 k

BT

ln(mB/m

A).

This

isno

longertrue

inthe

presenceofconstraints.

GR

OM

AC

Soffers

thepossibility

tointegrate

eq.3.77

oreq.3.78

inone

simulation

overthe

fullrange

fromA

toB

.H

owever,

ifthe

changeis

largeand

sampling

insufficiencycan

beexpected,

theuser

may

preferto

determine

thevalue

of〈dG/dλ〉

accuratelyat

anum

berof

well-chosen

intermediate

valuesofλ.

This

canbe

easilydone

bysetting

thestepsize

deltalam

bdato

zero.E

achsim

ulationcan

beequilibrated

first,anda

propererrorestimate

canbe

made

foreachvalue

ofdG/dλ

fromthe

fluctuationof∂H/∂λ

.T

hetotalfree

energychange

isthen

determined

afterwards

byan

appropriatenum

ericalintegrationprocedure.

Theλ

-dependencefor

theforce-field

contributionsis

describedin

sectionsec.

4.3.

3.13E

ssentialDynam

icsS

ampling

The

resultsfrom

EssentialD

ynamics

(seesec.8.9)

ofaprotein

canbe

usedto

guideM

Dsim

ula-tions.

The

ideais

thatfrom

aninitialM

Dsim

ulation(or

fromother

sources)a

definitionof

thecollective

fluctuationsw

ithlargestam

plitudeis

obtained.T

heposition

alongone

orm

oreofthese

collectivem

odescan

beconstrained

ina

(second)M

Dsim

ulationin

anum

berofw

aysfor

severalpurposes.

Forexam

ple,theposition

alonga

certainm

odem

aybe

keptfixedto

monitorthe

average

E.1

8.

gd

isre1

93

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-nin

de

x.nd

xInput,O

pt.Index

file-o

Mto

t.xvgO

utputxvgr/xm

grfile

-ee

psilo

n.xvg

Output

xvgr/xmgr

file-a

ave

r.xvgO

utputxvgr/xm

grfile

-dd

ipd

ist.xvgO

utputxvgr/xm

grfile

-cd

ipco

rr.xvgO

utput,Opt.

xvgr/xmgr

file-g

gkr.xvg

Output,O

pt.xvgr/xm

grfile

-qq

ua

dru

po

le.xvg

Output,O

pt.xvgr/xm

grfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-m

ureal

-1dipole

ofasingle

molecule

(inD

ebye)-m

um

ax

real5

max

dipolein

Debye

(forhistrogram

)-e

psilo

nR

Freal

0epsilon

ofthe

reactionfield

usedduring

thesim

ulation,needed

fordieclectric

constantcalculation.W

AR

NIN

G:0.0

means

infinity(default)

-skipint

0S

kipsteps

inthe

output(butnotinthe

computations)

-tem

preal

30

0average

temperature

ofthesim

ulation(needed

fordielectricconstantcal-

culation)-a

verco

rrbool

no

calculateA

Cfunction

ofaverage

dipolem

oment

ofthe

simulation

boxrather

thanaverage

ofAC

functionper

molecule

-gkra

tom

int0

Use

then-th

atomof

am

olecule(starting

from1)

tocalculate

thedis-

tancebetw

eenm

oleculesrather

thanthe

centerofcharge

(when

0)in

thecalculation

ofdistancedependentK

irkwood

factors-a

cflen

int-1

Lengthofthe

AC

F,defaultishalfthe

number

offrames

-no

rma

lizebool

yes

Norm

alizeA

CF

-Penum

0O

rderofLegendre

polynomialfor

AC

F(0

indicatesnone):

0,1

,2or3

-fitfnenum

no

ne

Fitfunction:n

on

e,e

xp,a

exp

,exp

exp

orvac

-ncskip

int0

Skip

Npoints

inthe

outputfileofcorrelation

functions-b

eg

infit

real0

Tim

ew

hereto

beginthe

exponentialfitofthecorrelation

function-e

nd

fitreal

-1T

ime

where

toend

theexponentialfitofthe

correlationfunction,-1

istill

theend

E.18

gdisre

gdisre

computes

violationsof

distancerestraints.

Ifnecessary

allprotons

canbe

addedto

aprotein

molecule.

The

programallw

ayscom

putesthe

instantaneousviolations

ratherthan

time-averaged,

becausethis

analysisis

donefrom

atrajectory

fileafterw

ardsitdoes

notmake

senseto

usetim

eaveraging.

An

indexfile

may

beused

toselectspecific

restraintsfor

printing.

Files

-sto

po

l.tpr

InputG

enericrun

input:tpr

tpbtpa

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-ds

drsu

m.xvg

Output

xvgr/xmgr

file-d

ad

rave

r.xvgO

utputxvgr/xm

grfile

-dn

drn

um

.xvgO

utputxvgr/xm

grfile

Page 57: CS USER AL Simulations - GROMACS

19

2A

pp

en

dix

E.

Ma

nu

alP

age

s

The

opth

erop

tion

isto

disc

retiz

eth

edi

hedr

alsp

ace

into

anu

mbe

rof

bins

,an

dgr

oup

each

conf

orm

atio

nin

dihe

dral

spac

ein

the

appr

opria

tebi

n.T

heou

tput

isth

engi

ven

asa

num

ber

ofdi

hedr

alco

nfor

mat

ions

sort

edac

cord

ing

tooc

cupa

ncy.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a-o

he

llo.o

ut

Out

put

Gen

eric

outp

utfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-s

abo

oln

oP

erfo

rmcl

uste

ran

alys

isin

dihe

dral

spac

ein

stea

dof

anal

ysin

gdi

hedr

altr

ansi

tions

.-m

ult

int

-1m

uliti

plic

ityfo

rdi

hedr

alan

gles

(by

defa

ultr

ead

from

topo

logy

)

•sh

ould

nota

skfo

rnu

mbe

rof

fram

es

E.1

7g

dipo

les

gdi

pole

sco

mpu

tes

the

tota

ldip

ole

plus

fluct

uatio

nsof

asi

mul

atio

nsy

stem

.F

rom

this

you

can

com

pute

e.g.

the

diel

ectr

icco

nsta

ntfo

rlo

wdi

elec

tric

med

ia

The

file

dip.

xvg

cont

ains

the

tota

ldip

ole

mom

ent

ofa

fram

e,th

eco

mpo

nent

sas

wel

las

the

norm

ofth

eve

ctor

.T

hefil

eav

er.x

vgco

ntai

ns<or

Muo

r2>

and<

orM

uor>

2du

ring

the

sim

ulat

ion.

The

file

dip.

xvg

cont

ains

the

dist

ribut

ion

ofdi

pole

mom

ents

durin

gth

esi

mul

atio

nT

hem

um

axis

used

asth

ehi

ghes

tval

uein

the

dist

ribut

ion

grap

h.

Fur

ther

mor

eth

edi

pole

auto

corr

elat

ion

func

tion

will

beco

mpu

ted,

whe

nop

tion

-cis

used

.It

can

beav

er-

aged

over

allm

olec

ules

,or

(with

optio

n-a

verc

orr)

itca

nbe

com

pute

das

the

auto

corr

elat

ion

ofth

eto

tal

dipo

lem

omen

toft

hesi

mul

atio

nbo

x.

Att

hem

omen

tthe

diel

ectr

icco

nsta

ntis

calc

ulat

edon

lyco

rrec

tifa

rect

angu

lar

orcu

bic

sim

ulat

ion

box

isus

ed.

Opt

ion

-gpr

oduc

esa

plot

ofth

edi

stan

cede

pend

ent

Kirk

woo

dG

-fac

tor,

asw

ella

sth

eav

erag

eco

sine

ofth

ean

gle

betw

een

the

dipo

les

asa

func

tion

ofth

edi

stan

ce.

The

plot

also

incl

udes

gOO

and

hOO

acco

rdin

gto

Nym

and

&Li

nse,

JCP

112

(200

0)pp

6386

-639

5.

EX

AM

PLE

S

gdi

pole

s-P

1-n

mol

s-o

dips

qr-m

u2.

273

-mum

ax5.

0-n

offt

Thi

sw

illca

lcul

ate

the

auto

corr

elat

ion

func

tion

ofth

em

olec

ular

dipo

les

usin

ga

first

orde

rLe

gend

repo

ly-

nom

ialo

fth

ean

gle

ofth

edi

pole

vect

oran

dits

elf

atim

et

late

r.F

orth

isca

lcul

atio

n10

01fr

ames

will

beus

ed.

Fur

ther

the

diel

ectr

icco

nsta

ntw

illbe

calc

ulat

edus

ing

anep

silo

nRF

ofin

finity

(def

ault)

,te

mpe

ra-

ture

of30

0K

(def

ault)

and

anav

erag

edi

pole

mom

ento

fthe

mol

ecul

eof

2.27

3(S

PC

).F

orth

edi

strib

utio

nfu

nctio

na

max

imum

of5.

0w

illbe

used

.

File

s -en

xe

ne

r.e

dr

Inpu

t,O

pt.

Gen

eric

ener

gy:

edr

ene

-ftr

aj.x

tcIn

put

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b

3.1

4.

Pa

ralle

liza

tion

37

forc

e(f

ree-

ener

gygr

adie

nt)

onth

atco

ordi

nate

inth

atpo

sitio

n.A

noth

erap

plic

atio

nis

toen

hanc

esa

mpl

ing

effic

ienc

yw

ithre

spec

tto

usua

lM

D[

34,

35].

Inth

isca

se,

the

syst

emis

enco

urag

edto

sam

ple

itsav

aila

ble

confi

gura

tion

spac

em

ore

syst

emat

ical

lyth

anin

adi

ffusi

on-li

kepa

thth

atpr

otei

nsus

ually

take

.

All

avai

labl

eco

nstr

aint

type

sar

ede

scrib

edin

the

appr

opria

tech

apte

rof

the

WH

ATIF

[36

]man

-ua

l.

3.14

Par

alle

lizat

ion

The

purp

ose

ofth

isse

ctio

nis

todi

scus

sth

epa

ralle

lizat

ion

ofth

epr

inci

ple

MD

algo

rithm

and

nott

ode

scrib

eth

eal

gorit

hms

that

are

inpr

actic

alus

efo

rmol

ecul

arsy

stem

sw

ithth

eirc

ompl

exva

riety

ofat

oms

and

term

sin

the

forc

efie

ldde

scrip

tions

.W

esh

allt

here

fore

cons

ider

asan

exam

ple

asi

mpl

esy

stem

cons

istin

gon

lyof

asi

ngle

type

ofat

oms

with

asi

mpl

efo

rmof

the

inte

ract

ion

pote

ntia

l.T

heem

phas

isw

illbe

onth

esp

ecia

lpro

blem

sth

atar

ise

whe

nth

eal

gorit

hmis

impl

emen

ted

ona

para

llelc

ompu

ter.

The

sim

ple

mod

elpr

oble

mal

read

yco

ntai

nsth

ebo

ttlen

eck

ofal

lM

Dsi

mul

atio

ns:

the

com

pu-

tatio

nally

inte

nsiv

eev

alua

tion

ofth

enon

-bo

nd

edf

orce

sbe

twee

npa

irsof

atom

s,ba

sed

onth

edi

stan

cebe

twee

npa

rtic

les.

Com

plex

mol

ecul

arsy

stem

sw

illin

addi

tion

invo

lve

man

ydi

ffere

ntki

nds

ofb

on

de

dfor

ces

betw

een

desi

gnat

edat

oms.

Suc

hin

tera

ctio

nsad

dto

the

com

plex

ityof

the

algo

rithm

butd

ono

tmod

ifyth

eba

sic

cons

ider

atio

nsco

ncer

ning

para

lleliz

atio

n.

3.14

.1M

etho

dsof

para

lleliz

atio

n

The

rear

ea

num

ber

ofm

etho

dsto

para

lleliz

eth

eM

Dal

gorit

hm,

each

ofth

emw

ithth

eir

own

adva

ntag

esan

ddi

sadv

anta

ges.

The

met

hod

toch

oose

depe

nds

onth

eha

rdw

are

and

com

pile

rsav

aila

ble.

We

listt

hem

here

:

1M

ess

age

Pa

ssin

g.

Inth

ism

etho

d,w

hich

ism

ore

orle

ssth

etr

aditi

onal

way

ofpa

ralle

lpro

gram

min

g,al

lthe

para

llelis

mis

expl

icitl

ypr

ogra

mm

edby

the

user

.T

hedi

sadv

anta

geis

that

itta

kes

extr

aco

dean

def

fort

,th

ead

vant

age

isth

atth

epr

ogra

mm

erke

eps

full

cont

rolo

ver

the

data

flow

and

can

doop

timiz

atio

nsa

com

pile

rco

uld

notc

ome

upw

ith.

The

impl

emen

tatio

nis

typi

cally

done

byca

lling

ase

tof

libra

ryro

utin

esto

send

and

re-

ceiv

eda

tato

and

from

othe

rpr

oces

sors

.A

lmos

tal

lhar

dwar

eve

ndor

ssu

ppor

tth

isw

ayof

para

llelis

min

thei

rC

and

For

tran

com

pile

rs.

2D

ata

Pa

ralle

l.T

his

met

hod

lets

the

user

defin

ear

rays

onw

hich

toop

erat

ein

para

llel.

Pro

gram

min

gth

isw

ayis

muc

hlik

eve

ctor

izin

g:re

curr

ence

isno

tpar

alle

lized

(e.g

.fo

r(i=

1;

(i<

MA

X);

i++

)a

[i]=

a[i-

1]

+1

;do

esno

tvec

toriz

ean

dno

tpar

alle

lize,

beca

use

for

ever

yit

here

sult

from

the

prev

ious

step

isne

eded

).

The

adva

ntag

eof

data

para

llelis

mis

that

itis

easi

erfo

rth

eus

er;t

heco

mpi

ler

take

sca

reof

the

para

llelis

m.

The

disa

dvan

tage

isth

atit

issu

ppor

ted

bya

smal

l(th

ough

grow

ing)

num

ber

Page 58: CS USER AL Simulations - GROMACS

38

Ch

ap

ter

3.

Alg

orith

ms

ofhardw

arevendors,

andthat

itis

much

harderto

maintain

aprogram

thathas

torun

onboth

parallelandsequentialm

achines,because

theonly

standardlanguage

thatsupports

itis

Fortran-90

which

isnotavailable

onm

anyplatform

s.

Both

methods

allowfor

theM

Dalgorithm

tobe

implem

entedw

ithoutm

uchtrouble.

Message

passingM

Dalgorithm

shave

beenpublished

sincethe

mid

80’s([

37],[38])

anddevelopm

entis

stillcontinuing.D

ataparallelprogram

ming

isnew

er,butstartingfrom

aw

ellvectorizedprogram

itisnothard

todo.

Our

implem

entationof

MD

isa

message

passingone,

thereason

forw

hichis

partlyhistorical:

theproject

todevelop

aparallel

MD

programstarted

when

Fortran-90

was

stillin

them

aking,and

nocom

pilersw

ereexpected

tobe

available.A

tcurrent,we

stillbelievethatm

essagepassing

isthe

way

togo,

afterhaving

donesom

eexperim

entsw

ithdata

parallelprogramm

ingon

aC

on-nection

Machine

(CM

-5),because

ofportability

toother

hardware,

thepoor

performance

ofthe

codeproduced

bythe

compilers

andbecause

thisw

ayofprogram

ming

hasthe

same

drawback

asvectorization:

thepartofthe

programthatis

notvectorizedor

parallelizeddeterm

inesthe

runtime

oftheprogram

(Am

dahl’slaw

).

The

approachw

etook

toparallelism

was

am

inimalistone:

useas

littlenon-standard

elements

inthe

software

aspossible,

anduse

thesim

plestprocessortopology

thatdoesthe

job.W

etherefore

decidedto

usea

standardlanguage

(AN

SI-C

)w

ithas

littlenon-standard

routinesas

possible.W

eonly

use5

comm

unicationroutines

thatarenon-standard.

Itistherefore

veryeasy

toportour

codeto

otherm

achines.

For

anO

(N2)

problemlike

MD

,one

ofthe

bestschem

esfor

theinterprocessor

connectionsis

aring,so

oursoftw

aredem

andsthata

ringis

presentinthe

interprocessorconnections.

Aring

canessentially

always

bem

appedonto

anothernetw

orklike

ahypercube,

abus

interface(E

thernete.g.

usingM

essageP

assingInterface

MP

I)or

atree

(CM

-5).S

ome

hardware

vendorshave

veryluxurious

connectionschem

esthat

connectevery

processorto

everyother

processor,but

we

donot

reallyneed

itand

sodo

notuse

iteven

thoughit

might

come

inhandy

attim

es.T

headvan-

tagew

iththis

simplistic

scheme

isthatG

RO

MA

CS

performs

extremely

welleven

oninexpensive

workstation

clusters.

When

usinga

message

passingschem

eone

hasto

dividethe

particlesover

processors,which

canbe

donein

two

ways:

•S

pa

ceD

eco

mp

ositio

n.

An

element

ofspace

isallocated

toeach

processor,w

hendividing

acubic

boxw

ithedge

boverP

processorsthis

canbe

doneby

givingeach

processora

slabof

lengthb/P

.T

hism

ethodhas

theadvantage

thateachprocessor

hasaboutthe

same

number

ofinteractionsto

calculate(at

leastw

henthe

simulated

systemhas

ahom

ogeneousdensity,

likea

liquidor

agas).

The

disadvantageis

thata

lotof

bookkeepingis

necessaryfor

particlesthat

move

overprocessor

boundaries.W

henusing

more

complex

systems

likem

acromolecules

thereare

also3-and

4-atominteractions

thatwould

complicate

thebookkeeping

som

uchthatthis

method

isnotused

inour

program.

•P

article

De

com

po

sition

.E

veryprocessor

isallocated

anum

berof

particles.W

hendividing

Nparticles

overPprocessors

eachprocessor

will

getN/P

particles.T

heim

plementation

ofthis

method

isdescribed

inthe

nextsection.

E.1

5.

gd

iele

ctric1

91

E.15

gdielectric

dielectriccalculates

frequencydependentdielectric

constantsfrom

theautocorrelation

functionofthe

totaldipole

mom

entin

yoursim

ulation.T

hisA

CF

canbe

generatedby

gdipoles.

For

anestim

ateof

theerror

youcan

rungstatistics

onthe

AC

F,and

usethe

outputthus

generatedfor

thisprogram

.T

hefunctional

forms

oftheavailable

functionsare:

One

parmeter

:y

=E

xp[-a1x]Tw

oparm

eters:

y=

a2E

xp[-a1x]T

hreeparm

eter:y

=a2

Exp[-a1

x]+(1

-a2)

Exp[-a3

x]Startvalues

forthe

fitprocedurecan

begiven

onthe

comm

andline.Itis

alsopossible

tofix

parameters

attheirstartvalue,use

-fixw

iththe

number

oftheparam

eteryou

wantto

fix.

Three

outputfilesare

generated,thefirstcontains

theA

CF,an

exponentialfittoitw

ith1,2

or3

parameters,

andthe

numericalderivative

ofthe

combination

data/fit.T

hesecond

filecontains

therealand

imaginary

partsof

thefrequency-dependent

dielectricconstant,

thelast

givesa

plotknow

nas

theC

ole-Cole

plot,in

which

theim

aginarycom

ponentis

plottedas

afunction

ofthe

realcom

ponent.F

ora

pureexponential

relaxation(D

ebyerelaxation)

thelatter

plotshouldbe

onehalfofa

circle

Files

-fM

tot.xvg

Inputxvgr/xm

grfile

-dd

eriv.xvg

Output

xvgr/xmgr

file-o

ep

sw.xvg

Output

xvgr/xmgr

file-c

cole

.xvgO

utputxvgr/xm

grfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-fft

booln

ouse

fastfouriertransform

forcorrelation

function-x1

boolye

suse

firstcolumn

asX

axisrather

thanfirstdata

set-e

int

real5

Tim

ew

ereto

endthe

integrationofthe

dataand

starttouse

thefit

-bfit

real5

Begin

time

offit-e

fitreal

50

0E

ndtim

eoffit

-tail

real5

00

Lengthoffunction

includingdata

andtailfrom

fit-A

real0

.5S

tartvaluefor

fitparameter

A-ta

u1

real1

0S

tartvaluefor

fitparameter

tau1-ta

u2

real1

Startvalue

forfitparam

etertau2

-ep

s0real

80

Epsilon

0ofyour

liquid-e

psR

Freal

78

.5E

psilonofthe

reactionfield

usedin

yoursim

ulation.A

valueof0

means

infinity.-fix

int0

Fix

parameters

attheirstartvalues,A

(2),tau1(1),or

tau2(4)

-ffnenum

no

ne

Fitfunction:n

on

e,e

xp,a

exp

,exp

exp

orvac

-nsm

oo

thint

3N

umber

ofpointsfor

smoothing

E.16

gdih

gdih

cando

two

things.T

hedefaultis

toanalyze

dihedraltransitionsby

merely

computing

allthedihedral

anglesdefined

inyour

topologyfor

thew

holetrajectory.

When

adihedralflips

overto

anotherm

inimum

anangle/tim

eplotis

made.

Page 59: CS USER AL Simulations - GROMACS

19

0A

pp

en

dix

E.

Ma

nu

alP

age

s

-lco

var.

log

Out

put

Log

file

-xp

mco

var.

xpm

Out

put,

Opt

.X

Pix

Map

com

patib

lem

atrix

file

-xp

ma

cova

ra.x

pm

Out

put,

Opt

.X

Pix

Map

com

patib

lem

atrix

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-tu

enum

ps

Tim

eun

it:p

s,f

s,n

s,u

s,m

s,s

,mor

h-f

itbo

olye

sF

itto

are

fere

nce

stru

ctur

e-r

ef

bool

no

Use

the

devi

atio

nfr

omth

eco

nfor

mat

ion

inth

est

ruct

ure

file

inst

ead

offr

omth

eav

erag

e-m

wa

bool

no

Mas

s-w

eigh

ted

cova

rianc

ean

alys

is-la

stin

t-1

Last

eige

nvec

tor

tow

rite

away

(-1

istil

lthe

last

)

E.1

4g

dens

ity

Com

pute

part

iald

ensi

ties

acro

ssth

ebo

x,us

ing

anin

dex

file.

Den

sitie

sin

gram

/cub

icce

ntim

eter

,nu

mbe

rde

nsiti

esor

elec

tron

dens

ities

can

beca

lcul

ated

.F

orel

ectr

onde

nsiti

es,e

ach

atom

isw

eigh

edby

itsat

omic

part

ialc

harg

e.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-nin

de

x.n

dx

Inpu

t,O

pt.

Inde

xfil

e-s

top

ol.t

pr

Inpu

tG

ener

icru

nin

put:

tpr

tpb

tpa

-ei

ele

ctro

ns.

da

tO

utpu

tG

ener

icda

tafil

e-o

de

nsi

ty.x

vgO

utpu

txv

gr/x

mgr

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-d

strin

gZ

Take

the

norm

alon

the

mem

bran

ein

dire

ctio

nX

,Yor

Z.

-sl

int

10

Div

ide

the

box

in#n

rsl

ices

.-n

um

be

rbo

oln

oC

alcu

late

num

ber

dens

ityin

stea

dof

mas

sde

nsity

.H

ydro

gens

are

not

coun

ted!

-ed

bool

no

Cal

cula

teel

ectr

onde

nsity

inst

ead

ofm

ass

dens

ity-c

ou

nt

bool

no

Onl

yco

unta

tom

sin

slic

es,n

ode

nsiti

es.

Hyd

roge

nsar

eno

tcou

nted

•W

hen

calc

ulat

ing

elec

tron

dens

ities

,ato

mna

mes

are

used

inst

ead

ofty

pes.

Thi

sis

bad.

•W

hen

calc

ulat

ing

num

ber

dens

ities

,at

oms

with

nam

esth

atst

art

with

Har

eno

tco

unte

d.T

his

may

besu

rpris

ing

ifyo

uus

ehy

drog

ens

with

nam

eslik

eO

P3.

3.1

4.

Pa

ralle

liza

tion

39

3.14

.2M

Don

arin

gof

proc

esso

rs

Whe

na

neig

hbor

listi

sno

tuse

dth

eM

Dpr

oble

mis

inpr

inci

ple

anO

(N2)p

robl

emas

each

part

icle

can

inte

ract

with

ever

yot

her.

Thi

sca

nbe

sim

plifi

edus

ing

New

ton’

sth

irdla

w

Fij

=−F

ji(3

.80)

Thi

sim

plie

sth

atth

ere

isha

lfa

mat

rixof

inte

ract

ions

(with

outd

iago

nal,

apa

rtic

ledo

esno

tint

erac

tw

ithits

elf)

toco

nsid

er(F

ig.3.

10).

Whe

nw

ere

flect

the

uppe

rrig

httr

iang

leof

inte

ract

ions

toth

elo

wer

left

tria

ngle

ofth

em

atrix

,w

est

illco

ver

all

poss

ible

inte

ract

ions

,bu

tno

wev

ery

row

inth

em

atrix

has

alm

ost

the

sam

enu

mbe

rof

poin

tsor

poss

ible

inte

ract

ions

.W

eca

nno

was

sign

a(p

refe

rabl

yeq

ual)

num

ber

ofro

ws

toea

chpr

oces

sor

toco

mpu

teth

efo

rces

and

atth

esa

me

time

anu

mbe

rof

part

icle

sto

doth

eup

date

on,

the

ho

me

part

icle

s.T

henu

mbe

rof

inte

ract

ions

per

part

icle

isde

pend

ento

nth

etota

lnu

mb

erN

ofpa

rtic

les

(see

Fig

.3.11

)an

don

thep

art

icle

nu

mb

er

i.T

heex

actf

orm

ulae

are

give

nin

Tabl

e3.

2.

Aflo

wch

arto

fthe

algo

rithm

isgi

ven

inF

ig.3.12

.

Itis

the

sam

eas

the

sequ

entia

lalg

orith

m,e

xcep

tfor

two

com

mun

icat

ion

step

s.A

fter

the

part

icle

sha

vebe

enre

set

inth

ebo

x,ea

chpr

oces

sor

send

sits

coor

dina

tes

left

and

then

star

tsco

mpu

tatio

nof

the

forc

es.

Afte

rth

isst

epea

chpr

oces

sor

hold

sth

ep

art

ialf

orc

esf

orth

eav

aila

ble

part

icle

s,e.

g.pr

oces

sor

0ho

lds

forc

esac

ting

onho

me

part

icle

sfr

ompr

oces

sor

0,1,

2an

d3.

The

sefo

rces

mus

tbe

accu

mul

ated

and

sent

back

(rig

ht)

toth

eho

me

proc

esso

r.F

inal

lyth

eup

date

ofth

eve

loci

tyan

dco

ordi

nate

sis

done

onth

eho

me

proc

esso

r.

The

com

mu

nic

ate

rro

utin

eis

give

nbe

low

inth

efu

llC

-cod

e:

void

com

mu

nic

ate

_r(

int

np

rocs

,int

pid

,rve

cve

cs[]

,int

sta

rt[]

,int

ho

me

nr[

])/*

*n

pro

cs=

nu

mb

er

of

pro

cess

ors

*p

id=

pro

cess

or

id(0

..n

pro

cs-1

)*

vecs

=ve

cto

rs*

sta

rt=

sta

rtin

gin

de

xin

vecs

for

ea

chp

roce

sso

r*

ho

me

nr

=n

um

be

ro

fh

om

ep

art

icle

sfo

re

ach

pro

cess

or

*/{

int

i;/*

pro

cess

or

cou

nte

r*/

int

shift

;/*

the

am

ou

nt

of

pro

cess

ors

toco

mm

un

ica

tew

ith*/

int

cur;

/*cu

rre

nt

pro

cess

or

tose

nd

da

tafr

om

*/in

tn

ext

;/*

ne

xtp

roce

sso

ro

na

rin

g(u

sin

gm

od

ulo

)*/

cur

=p

id;

shift

=n

pro

cs/2

;

for

(i=

0;

(i<

shift

);i+

+)

{n

ext

=(c

ur+

1)

%n

pro

cs;

sen

d(le

ft,

vecs

[sta

rt[c

ur]

],h

om

en

r[cu

r]);

rece

ive

(rig

ht,

vecs

[sta

rt[n

ext

]],

ho

me

nr[

ne

xt])

;cu

r=n

ext

;}

} The

data

flow

arou

ndth

erin

gis

visu

aliz

edin

Fig

.3.

13.

Not

eth

atbe

caus

eof

the

ring

topo

logy

each

proc

esso

rau

tom

atic

ally

gets

the

prop

erpa

rtic

les

toin

tera

ctw

ith.

Page 60: CS USER AL Simulations - GROMACS

40

Ch

ap

ter

3.

Alg

orith

ms

012345678

012345678

01

23

45

67

80

12

34

56

78

jj

ii

Figure

3.10:T

heinteraction

matrix

(left)and

thesam

eusing

action=

−reaction

(right).

imod

2=

0im

od2

=0

imod

2=

1im

od2

=1

i<

N/2

i≥N

/2i<

N/2

i≥N

/2N

mod

2=

1N/2

N/2

N/2

N/2

Nm

od4

=2

N/2

N/2

N/2−

1N/2−

1N

mod

4=

0N/2

N/2−

1N/2−

1N/2

Table3.2:

The

number

ofinteractionsbetw

eenparticles.

The

number

ofj

particlesperi

particleis

afunction

ofthetotalnum

berofparticlesN

andparticle

numberi.

Note

thatherethe/

operatoris

usedfor

integerdivision,i.e.truncating

therem

inder.

j

i

j

i

j

i012345

01

23

45

01234567

01

23

45

67

0123456

01

23

45

6

N m

od 4 = 2N

mod 2 = 1

N m

od 4 = 0

Figure

3.11:Interaction

matrices

fordifferent

N.

The

number

ofj-particlesani-particle

interactsw

ithdepends

ontheto

talnum

berofparticles

andon

theparticle

nu

mb

er.

E.1

2.

gco

nfrm

s1

89

-see

dint

19

93

Random

number

seedfor

Monte

Carlo

clusteringalgorithm

-nite

rint

10

00

0N

umber

ofiterationsfor

MC

-kTreal

0.0

01

Boltzm

annw

eightingfactor

forM

onteC

arlooptim

ization(zero

turnsoff

uphillsteps)

E.12

gconfrm

s

gconfrm

scom

putesthe

rootmean

squaredeviation

(RM

SD

)oftw

ostructures

afterLS

Qfitting

thesecond

structureon

thefirstone.

The

two

structuresdo

NO

Tneed

tohave

thesam

enum

berofatom

s,onlythe

two

indexgroups

usedfor

thefitneed

tobe

identical.

The

superimposed

structuresare

written

tofile.

Ina

.pd

bfile

thetw

ostructures

willbe

written

asseparate

models

(userasm

ol

-nm

rpd

b).

Files-f1

con

f1.g

roInput

Structure+

mass(db):

tprtpb

tpagro

g96pdb

-f2co

nf2

.gro

InputG

enericstructure:

grog96

pdbtpr

tpbtpa

-ofit.p

db

Output

Generic

structure:gro

g96pdb

-n1

fit1.n

dx

Input,Opt.

Indexfile

-n2

fit2.n

dx

Input,Opt.

Indexfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-on

ebool

no

Only

write

thefitted

structureto

file-p

bc

booln

oT

ryto

make

molecules

whole

again

E.13

gcovar

gco

var

calculatesand

diagonalizesthe

(mass-w

eighted)covariance

matrix.

Allstructures

arefitted

tothe

structurein

thestructure

file.W

henthis

isnota

runinputfile

periodicityw

illnotbetaken

intoaccount.

When

thefit

andanalysis

groupsare

identicalandthe

analysisis

nonm

ass-weighted,

thefit

willalso

benon

mass-w

eighted.

The

eigenvectorsare

written

toa

trajectoryfile

(-v

).W

henthe

same

atoms

areused

forthe

fitand

thecovariance

analysis,the

referencestructure

forthe

fitis

written

firstw

itht=

-1.T

heaverage

(orreference

when

-ref

isused)structure

isw

rittenw

itht=

0,theeigenvectors

arew

rittenas

frames

with

theeigenvector

number

astim

estamp.

The

eigenvectorscan

beanalyzed

with

ga

na

eig

.

Option

-xpm

writes

thew

holecovariance

matrix

toan

xpmfile.

Option

-xpm

aw

ritesthe

atomic

covariancem

atrixto

anxpm

file,i.e.

foreach

atompair

thesum

ofthe

xx,yyand

zzcovariances

isw

ritten.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-n

ind

ex.n

dx

Input,Opt.

Indexfile

-oe

ige

nva

l.xvgO

utputxvgr/xm

grfile

-ve

ige

nve

c.trrO

utputF

ullprecisiontrajectory:

trrtrj

-av

ave

rag

e.p

db

Output

Generic

structure:gro

g96pdb

Page 61: CS USER AL Simulations - GROMACS

18

8A

pp

en

dix

E.

Ma

nu

alP

age

s

the

low

errig

htha

lf(d

epen

dson-

ma

xan

d-k

ee

pfr

ee

).-g

writ

esin

form

atio

non

the

optio

nsus

edan

da

deta

iled

listo

fall

clus

ters

and

thei

rm

embe

rs.

Add

ition

ally

,anu

mbe

rof

optio

nalo

utpu

tfile

sca

nbe

writ

ten:

-dis

tw

rites

the

RM

SD

dist

ribut

ion.

-ev

writ

esth

eei

genv

ecto

rsof

the

RM

SD

mat

rixdi

agon

aliz

atio

n.-s

zw

rites

the

clus

ter

size

s.-t

rw

rites

am

atrix

ofth

enu

mbe

rtr

ansi

tions

betw

een

clus

ter

pairs

.-n

trw

rites

the

tota

lnum

ber

oftr

ansi

tions

toor

from

each

clus

ter.

-clid

writ

esth

ecl

uste

rnu

mbe

ras

afu

nctio

nof

time.

-cl

writ

esav

erag

e(w

ithop

tion-a

v)

orce

ntra

lst

ruct

ure

ofea

chcl

uste

ror

writ

esnu

mbe

red

files

with

clus

ter

mem

bers

for

ase

lect

edse

tofc

lust

ers

(with

optio

n-w

cl,d

epen

dson

-nst

and

-rm

smin

).

File

s-f

tra

j.xtc

Inpu

t,O

pt.

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b-s

top

ol.t

pr

Inpu

t,O

pt.

Str

uctu

re+

mas

s(db

):tp

rtp

btp

agr

og9

6pd

b-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-dm

rmsd

.xp

mIn

put,

Opt

.X

Pix

Map

com

patib

lem

atrix

file

-orm

sd-c

lust

.xp

mO

utpu

tX

Pix

Map

com

patib

lem

atrix

file

-gcl

ust

er.

log

Out

put

Log

file

-dis

trm

sd-d

ist.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e-e

vrm

sd-e

ig.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-s

zcl

ust

-siz

e.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-t

rclu

st-t

ran

s.xp

mO

utpu

t,O

pt.

XP

ixM

apco

mpa

tible

mat

rixfil

e-n

trclu

st-t

ran

s.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e-c

lidcl

ust

-id

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-cl

clu

ste

rs.p

db

Out

put,

Opt

.G

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-tu

enum

ps

Tim

eun

it:p

s,f

s,n

s,u

s,m

s,s

,mor

h-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-dis

tabo

oln

oU

seR

MS

Dof

dist

ance

sin

stea

dof

RM

Sde

viat

ion

-nle

vels

int

40

Dis

cret

ize

RM

SD

mat

rixin

#le

vels

-ke

ep

fre

ein

t-4

if>

0#

leve

lsno

tto

use

whe

nco

lorin

gcl

uste

rs;i

f<

0nl

evel

s/-k

eepf

ree+

1le

vels

will

notb

eus

ed-c

uto

ffre

al0

.1R

MS

Dcu

t-of

f(nm

)fo

rtw

ost

ruct

ures

tobe

neig

hbor

-ma

xre

al-1

Max

imum

leve

lin

RM

SD

mat

rix-s

kip

int

1O

nly

anal

yze

ever

ynr

-th

fram

e-a

vbo

oln

oW

rite

aver

age

iso

mid

dle

stru

ctur

efo

rea

chcl

uste

r-w

clin

t0

Writ

eal

lstr

uctu

res

for

first

#cl

uste

rsto

num

bere

dfil

es-n

stin

t1

Onl

yw

rite

alls

truc

ture

sif

mor

eth

an#

per

clus

ter

-rm

smin

real

0m

inim

umrm

sdi

ffere

nce

with

rest

ofcl

uste

rfo

rw

ritin

gst

ruct

ures

-me

tho

den

umlin

kag

eM

etho

dfo

rcl

uste

rde

term

inat

ion:

linka

ge

,ja

rvis

-pa

tric

k,

mo

nte

-ca

rlo

,dia

go

na

liza

tion

org

rom

os

-bin

ary

bool

no

Tre

atth

eR

MS

Dm

atrix

asco

nsis

ting

of0

and

1,w

here

the

cut-

off

isgi

ven

by-c

utof

f-M

int

10

Num

ber

ofne

ares

tne

ighb

ors

cons

ider

edfo

rJa

rvis

-Pat

rick

algo

rithm

,0

isus

ecu

toff

-Pin

t3

Num

ber

ofid

entic

alne

ares

tnei

ghbo

rsre

quire

dto

form

acl

uste

r

3.1

4.

Pa

ralle

liza

tion

41

read

_dat

a

Don

e

NO

outp

ut_s

tep

upda

te_r

_and

_v

mor

e st

eps

?Y

ES

com

pute

_for

ces

* *

rese

t_r_

in_b

ox

com

mun

icat

e_r

com

mun

icat

e_an

d_su

m_f

Fig

ure

3.12

:T

heP

aral

lelM

Dal

gorit

hm.

Ifth

est

eps

mar

ked

*ar

ele

ftou

twe

have

the

sequ

entia

lal

gorit

hmag

ain.

Page 62: CS USER AL Simulations - GROMACS

42

Ch

ap

ter

3.

Alg

orith

ms

01

2

34

5

6

7Forces

Coordinates

Figure

3.13:D

ataflow

ina

ringofprocessors.

3.15P

arallelMolecular

Dynam

ics

Inthis

chapterw

edescribe

some

detailsof

theparallelM

Dalgorithm

usedin

GR

OM

AC

S.

This

alsoincludes

some

otherinform

ationon

neighborsearching

anda

sideexcursion

toparallelsort-

ing.P

leasenote

thefollow

ingw

hichw

euse

throughoutthischapter:

definition:N

:N

umber

ofparticles,Mnum

berofprocessors.

GR

OM

AC

Sem

ploystw

odifferentgrids:

theneighbor

searchinggrid

(NS

grid)and

thecom

binedcharge/potentialgrid

(FF

Tgrid),

asw

illbedescribed

below.

Tom

aximize

theconfusion,

thesetw

ogrids

arem

appedonto

agrid

ofprocessorsw

henG

RO

MA

CS

runson

aparallelcom

puter.

3.15.1D

omain

decomposition

Modern

dayparallelcom

puters,such

asan

IBM

SP

/2or

aC

rayT

3Econsist

ofrelatively

small

numbers

ofrelativelyfastscalar

processors(typically

8to

256).T

hecom

munication

channelsthat

areavailable

inhardw

areon

thesem

achineare

notdirectlyvisible

forthe

programm

er,asoftw

arelayer(usually

MP

I)hidesthis,and

makes

comm

unicationfrom

allprocessorsto

allotherspossible.

Incontrast,in

theG

RO

MA

CS

hardware

[1]only

comm

unicationin

aring

was

available,i.e.eachprocessor

couldcom

municate

with

itsdirectneighbors

only.

Itseem

slogicalto

map

thecom

putationalboxof

anM

Dsim

ulationsystem

toa

3Dgrid

ofpro-

cessors(e.g.4x4x4

fora

64processor

system).

This

ensuresthatm

ostinteractionsthatare

localinspace

canbe

computed

with

information

fromneighboring

processorsonly.

How

ever,thism

eansthat

therehave

tobe

comm

unicationchannels

in3

dimensions

too,w

hichis

notnecessarily

thecase.

Although

thism

aybe

overcome

insoftw

are,such

am

appingis

complicated

forthe

MD

software

asw

ell,withoutclear

benefitsin

terms

ofperformance

form

ostparallelcomputers.

Therefore

we

optfora

simple

one-dimensionaldivision

scheme

forthe

computationalbox.

Each

processorgets

aslab

ofthisbox

inthe

X-dim

ension.F

orthe

comm

unicationbetw

eenprocessors

thishas

two

main

advantages:

1.S

implicity

ofcoding.C

omm

unicationcan

onlybe

totw

oneighbors

(calledle

ftandrig

htin

GR

OM

AC

S).

E.1

1.

gclu

ster

18

7

-om

eg

abool

no

Outputfor

Om

egadihedrals

(peptidebonds)

-ram

abool

no

Generate

Phi/P

siandC

hi1/Chi2

ramachandran

plots-vio

lbool

no

Write

afile

thatgives0

or1

forviolated

Ram

achandranangles

-all

booln

oO

utputseparatefiles

forevery

dihedral.-sh

iftbool

no

Com

putechem

icalshiftsfrom

Phi/P

siangles-ru

nint

1perform

runningaverage

overndeg

degreesfor

histograms

-ma

xchi

enum0

calculatefirstndih

Chidihedrals:0,1

,2,3

,4,5

or6-n

orm

histo

boolye

sN

ormalize

histograms

-ram

om

eg

abool

no

compute

averageom

egaas

afunction

ofphi/psiandplotitin

anxpm

plot-b

fact

real-1

B-factor

valuefor

pdbfile

foratom

sw

ithno

calculateddihedral

orderparam

eter-b

ma

xreal

0M

aximum

B-factor

onany

oftheatom

sthatm

akeup

adihedral,

forthe

dihedralangleto

beconsidere

inthe

statistics.A

ppliesto

databasew

orkw

herea

number

ofX-R

aystructures

isanalyzed.

-bmax

<=

0m

eansno

limit.

-acfle

nint

-1Length

oftheA

CF,defaultis

halfthenum

beroffram

es-n

orm

alize

boolye

sN

ormalize

AC

F-P

enum0

Order

ofLegendrepolynom

ialforA

CF

(0indicates

none):0

,1,2

or3-fitfn

enumn

on

eF

itfunction:no

ne

,exp

,ae

xp,e

xpe

xporva

c-n

cskipint

0S

kipN

pointsin

theoutputfile

ofcorrelationfunctions

-be

gin

fitreal

0T

ime

where

tobegin

theexponentialfitofthe

correlationfunction

-en

dfit

real-1

Tim

ew

hereto

endthe

exponentialfitofthecorrelation

function,-1is

tillthe

end

•P

roducesM

AN

Youtputfiles

(upto

about4tim

esthe

number

ofresiduesin

theprotein,tw

icethatif

autocorrelationfunctions

arecalculated).

Typicallyseveralhundred

filesare

output.

E.11

gcluster

gcluster

cancluster

structuresw

ithseveraldifferent

methods.

Distances

between

structurescan

bedeter-

mined

froma

trajectoryor

readfrom

anX

PM

matrix

filew

iththe-d

moption.

RM

Sdeviation

afterfitting

orR

MS

deviationofatom

-pairdistances

canbe

usedto

definethe

distancebetw

eenstructures.

fulllinkage:add

astructure

toa

clusterw

henits

distanceto

anyelem

entofthecluster

isless

thancu

toff

.

JarvisP

atrick:add

astructure

toa

clusterw

henthis

structureand

astructure

inthe

clusterhave

eachother

asneighbors

andthey

havea

leastP

neighborsin

comm

on.T

heneighbors

ofastructure

arethe

Mclosest

structuresor

allstructuresw

ithincuto

ff.

Monte

Carlo:

reorderthe

RM

SD

matrix

usingM

onteC

arlo.

diagonalization:diagonalize

theR

MS

Dm

atrix.

gromos:

usealgorithm

asdescribed

inD

aurae

tal.(A

ngew

.C

he

m.

Int.

Ed

.1999,38,pp

236-240).C

ountnum

berofneighbors

usingcut-off,take

structurew

ithlargestnum

berofneighbors

with

allitsneighbors

ascluster

andelem

inateitfrom

thepoolofclusters.

Repeatfor

remaining

structuresin

pool.

When

theclustering

algorithmassigns

eachstructure

toexactly

onecluster

(fulllinkage,JarvisP

atrickand

gromos)

anda

trajectoryfile

issupplied,

thestructure

with

thesm

allestaverage

distanceto

theothers

orthe

averagestructure

orallstructures

foreach

clusterw

illbew

rittento

atrajectory

file.W

henw

ritingall

structures,separatenum

beredfiles

arem

adefor

eachcluster.

Two

outputfilesare

always

written:

-ow

ritesthe

RM

SD

valuesin

theupper

lefthalfofthem

atrixand

agraphicaldepiction

oftheclusters

in

Page 63: CS USER AL Simulations - GROMACS

18

6A

pp

en

dix

E.

Ma

nu

alP

age

s

-otr

bu

ntil

tr.x

vgO

utpu

txv

gr/x

mgr

file

-otl

bu

ntil

tl.xv

gO

utpu

txv

gr/x

mgr

file

-ok

bu

nki

nk.

xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-okr

bu

nki

nkr

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-okl

bu

nki

nkl

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-oa

axe

s.p

db

Out

put,

Opt

.P

rote

inda

taba

nkfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-tu

enum

ps

Tim

eun

it:p

s,f

s,n

s,u

s,m

s,s

,mor

h-n

ain

t0

Num

ber

ofax

es-z

bool

no

Use

the

Z-a

xis

asre

fere

nce

iso

the

aver

age

axis

E.1

0g

chi

gch

icom

pute

sph

i,ps

i,om

ega

and

chid

ihed

rals

for

ally

our

amin

oac

idba

ckbo

nean

dsi

dech

ains

.It

can

com

pute

dihe

dral

angl

eas

afu

nctio

nof

time,

and

ashi

stog

ram

dist

ribut

ions

.O

utpu

tis

info

rmof

xvgr

files

,as

wel

las

aLa

TeX

tabl

eof

the

num

ber

oftr

ansi

tions

per

nano

seco

nd.

Ord

erpa

ram

eter

sS

2fo

rea

chof

the

dihe

dral

sar

eca

lcul

ated

and

outp

utas

xvgr

file

and

optio

nally

asa

pdb

file

with

the

S2

valu

esas

B-f

acto

r.

Ifop

tion

-cis

give

n,th

epr

ogra

mw

illca

lcul

ate

dihe

dral

auto

corr

elat

ion

func

tions

.T

hefu

nctio

nus

edis

C(t

)=<

cos(

chi(t

au))

cos(

chi(t

au+

t))>.

The

use

ofco

sine

sra

ther

than

angl

esth

emse

lves

,re

solv

esth

epr

oble

mof

perio

dici

ty.

(Van

der

Spo

el&

Ber

ends

en(1

997)

,B

ioph

ys.

J.72

,203

2-20

41).

The

optio

n-r

gene

rate

sa

cont

our

plot

ofth

eav

erag

eom

ega

angl

eas

afu

nctio

nof

the

phia

ndps

iang

les,

that

is,i

na

Ram

acha

ndra

npl

otth

eav

erag

eom

ega

angl

eis

plot

ted

usin

gco

lor

codi

ng.

File

s-c

con

f.g

roIn

put

Gen

eric

stru

ctur

e:gr

og9

6pd

btp

rtp

btp

a-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-oo

rde

r.xv

gO

utpu

txv

gr/x

mgr

file

-po

rde

r.p

db

Out

put,

Opt

.P

rote

inda

taba

nkfil

e-s

sss

du

mp

.da

tIn

put,

Opt

.G

ener

icda

tafil

e-jc

Jco

up

ling

.xvg

Out

put

xvgr

/xm

grfil

e-c

orr

dih

corr

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-gch

i.lo

gO

utpu

tLo

gfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-r

0in

t1

star

ting

resi

due

-ph

ibo

oln

oO

utpu

tfor

Phi

dihe

dral

angl

es-p

sibo

oln

oO

utpu

tfor

Psi

dihe

dral

angl

es

3.1

5.

Pa

ralle

lMo

lecu

lar

Dyn

am

ics

43

X Y Z0

12

N-1

01

23

45

CPU

num

ber

atom

num

ber

inde

x

coor

d.

Fig

ure

3.14

:In

dex

inth

eco

ordi

nate

arra

y.T

hedi

visi

onin

slab

sis

indi

cate

dby

dash

edlin

es.

2.C

omm

unic

atio

nca

nus

ually

bedo

nein

larg

ech

unks

,whi

chm

akes

itm

ore

effic

ient

onm

ost

hard

war

epl

atfo

rms.

Mos

tint

erac

tions

inm

olec

ular

dyna

mic

sha

vein

prin

cipl

ea

shor

tran

ged

char

acte

r.B

onds

,ang

les

and

dihe

dral

sar

egu

aran

teed

toha

veth

eco

rres

pond

ing

part

icle

scl

ose

insp

ace.

3.15

.2D

omai

nde

com

posi

tion

for

non-

bond

edfo

rces

For

larg

epa

ralle

lco

mpu

ters

,do

mai

nde

com

posi

tion

ispr

efer

able

over

part

icle

deco

mpo

sitio

n,si

nce

itis

easi

erto

dolo

adba

lanc

ing.

With

out

load

bala

ncin

gth

esc

alin

gof

the

code

isra

ther

poor

...F

orth

ispu

rpos

e,th

eco

mpu

tatio

nalb

oxis

divi

ded

inM

slab

s,w

hereM

iseq

ualt

oth

enu

mbe

rof

proc

esso

rs.

The

rear

em

ultip

lew

ays

ofdi

vidi

ngth

ebo

xov

erpr

oces

sors

,but

sinc

eth

eG

RO

MA

CS

code

assu

mes

arin

gto

polo

gyfo

rthe

proc

esso

rs,i

tis

logi

calt

ocu

tthe

syst

emin

slab

sin

just

one

dim

ensi

on,t

heX

dim

ensi

on.

The

algo

rithm

for

neig

hbor

sear

chin

gth

enbe

com

es:

1.M

ake

alis

tof

char

gegr

oup

indi

ces

sort

edon

(incr

easi

ng)

Xco

ordi

nate

(Fig

.3.

14).

Not

eth

atca

rem

ustb

eta

ken

topa

ralle

lize

the

sort

ing

algo

rithm

asw

ell.

See

sec.

3.15

.4.

2.D

ivid

eth

islis

tint

osl

abs,

such

that

each

slab

has

the

sam

enu

mbe

rof

char

gegr

oups

3.P

utth

epa

rtic

les

corr

espo

ndin

gto

the

loca

lsla

bon

a3D

NS

grid

asde

scrib

edin

sec.

3.4.

2.

4.C

omm

unic

ate

the

NS

grid

tone

ighb

orin

gpr

oces

sors

(not

nece

ssar

ilyto

allp

roce

ssor

s).

The

amou

ntof

neig

hbor

ing

NS

grid

cells

(N gx)

toco

mm

unic

ate

isde

term

ined

byth

ecu

t-of

fle

ngthr c

acco

rdin

gto

Ngx

=r cM l x

(3.8

1)

whe

rel x

isth

ebo

xle

ngth

inth

esl

abbi

ngdi

rect

ion.

5.O

nea

chpr

oces

sor

com

pute

the

neig

hbor

list

for

all

char

gegr

oups

inits

slab

usin

gth

eno

rmal

grid

neig

hbor

-sea

rchi

ng.

For

hom

ogen

eous

syst

em,

this

iscl

ose

toan

optim

allo

adba

lanc

ing,

with

out

actu

ally

doin

glo

adba

lanc

ing.

For

inho

mog

eneo

ussy

stem

,suc

has

mem

bran

es,o

rin

terf

aces

,the

dim

ensi

onfo

rsl

ab-

bing

mus

tbe

chos

ensu

chth

atit

ispe

rpen

dicu

lart

oth

ein

terf

ace;

inth

isfa

shio

nea

chpr

oces

sorh

as

Page 64: CS USER AL Simulations - GROMACS

44

Ch

ap

ter

3.

Alg

orith

ms

“alittle

bitofeverything”.T

heG

RO

MA

CS

utilityprograme

ditco

nf

hasan

optionto

rotatea

whole

computationalbox.

The

following

observationsare

importanthere:

•P

articlesm

aydiffuse

fromone

slabto

theother,therefore

eachprocessor

musthold

coordi-nates

forallparticles

allthetim

e,anddistribute

forcesback

toallprocessors

asw

ell.

•Velocities

arekepton

the“hom

eprocessor”

foreach

particle,where

theintegration

ofNew

-ton’s

equationsis

done.

•F

ixedinteraction

lists(bonds,

anglesetc.)

arekept

eachon

asingle

processor.S

inceall

processorshave

allcoordinates,it

doesnot

matter

where

interactionsare

calculated.T

hedivision

isactually

doneby

theG

RO

MA

CS

preprocessorg

rom

pp

andcare

istaken

that,asfar

aspossible,every

processorgets

thesam

enum

berofbonded

interactions.

Inall,

thism

akesfor

am

ixedparticle

decomposition/dom

aindecom

positionschem

efor

paral-lelization

ofthe

MD

code.T

hecom

munication

costsare

fourtim

eshigher

thanfor

thesim

pleparticle

decomposition

method

describedin

sec.3.14

(thew

holecoordinate

andforce

arrayare

comm

unicatedacross

thew

holering,

ratherthan

halfthe

arrayover

halfthe

ring).H

owever,

forlarge

numbers

ofprocessorsthe

improved

loadbalancing

compensates

thiseasily.

3.15.3P

arallelPP

PM

Afurther

reasonfor

domain

decomposition

isthe

PP

PM

algorithm.

This

algorithmw

orksw

itha

3DF

astF

ourierT

ransform.

Item

ploysa

discretegrid

ofdim

ensions(

nx ,n

y ,nz ),

theF

FT

grid.T

healgorithm

consistoffivesteps,each

ofwhich

haveto

beparallelized:

1.S

preadingcharges

onthe

FF

Tgrid

toobtain

thecharge

distributionρ(r).

This

bitinvolvesthe

following

sub-steps:

a.putparticle

inthe

box

b.find

theF

FT

gridcellin

which

theparticle

resides

c.add

thecharge

ofthe

particletim

esthe

appropriatew

eightfactor

(seesec.

4.6.3)to

eachofthe

27grid

points(3

x3

x3).

Inthe

parallelcase,theF

FT

gridm

ustbefilled

oneach

processorw

ithits

shareofthe

par-ticles,

andsubsequently

theF

FT

gridsof

allprocessorsm

ustbe

summ

edto

findthe

totalcharge

distribution.It

may

beclear

thatthis

inducesa

largeam

ountof

unnecessaryw

ork,unless

we

usedom

aindecom

position.If

eachprocessor

onlyhas

particlesin

acertain

re-gion

ofspace,itonlyhas

tocalculate

thecharge

distributionfor

thatregionofspace.

Since

GR

OM

AC

Sw

orksw

ithslabs,this

means

thateachprocessor

fillsthe

FF

Tgrid

cellscorre-

spondingto

it’sslab

inspace

andaddition

ofFF

Tgrids

needonly

bedone

forneighboring

slabs.To

bem

oreprecise,the

slabxforprocessori

isdefined

as:

ilxM

≤x<

(i+1)lxM

(3.82)

E.8

.g

bo

nd

18

5

E.8

gbond

gbond

makes

adistribution

ofbond

lengths.If

allisw

ellagaussian

distributionshould

bem

adew

henusing

aharm

onicpotential.

bondsare

readfrom

asingle

groupin

theindex

filein

orderi1-j1

i2-j2thru

in-jn.

-tol

givesthe

half-width

ofthe

distributionas

afraction

ofthe

bondlength(

-ble

n).

That

means,

fora

bondof0.2

atolof0.1

givesa

distributionfrom

0.18to

0.22

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-nin

de

x.nd

xInput

Indexfile

-ob

on

ds.xvg

Output

xvgr/xmgr

file-l

bo

nd

s.log

Output,O

pt.Log

file

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-btim

e-1

Firstfram

e(ps)

toread

fromtrajectory

-etim

e-1

Lastframe

(ps)to

readfrom

trajectory-d

ttim

e-1

Only

usefram

ew

hentM

OD

dt=firsttim

e(ps)

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-b

len

real-1

Bond

length.B

ydefaultlength

offirstbond-to

lreal

0.1

Halfw

idthofdistribution

asfraction

ofblen-a

ver

boolye

sS

umup

distributions

•Itshould

bepossible

togetbond

information

fromthe

topology.

E.9

gbundle

gbundle

analyzesbundles

ofaxes.T

heaxes

canbe

forinstance

helixaxes.

The

programreads

two

indexgroups

anddivides

bothofthem

in-naparts.

The

centersofm

assofthese

partsdefine

thetops

andbottom

sof

theaxes.

Severalquantities

arew

rittento

file:the

axislength,

thedistance

andthe

z-shiftof

theaxis

mid-points

with

respecttothe

averagecenter

ofallaxes,thetotaltilt,the

radialtiltandthe

lateraltiltwith

respecttothe

averageaxis.

With

options-ok

,-okr

and-o

klthe

total,radialandlateralkinks

oftheaxes

areplotted.

An

extraindex

groupofkink

atoms

isrequired,w

hichis

alsodivided

into-n

aparts.

The

kinkangle

isdefined

asthe

anglebetw

eenthe

kink-topand

thebottom

-kinkvectors.

With

option-o

athe

top,m

id(or

kinkw

hen-ok

isset)

andbottom

pointsof

eachaxis

arew

rittento

apdb

fileeach

frame.

The

residuenum

berscorrespond

tothe

axisnum

bers.W

henview

ingthis

filew

ithra

smo

l,

usethe

comm

andline

option-nm

rpd

b,

andtypese

ta

xistru

eto

displaythe

referenceaxis.

Files

-ftra

j.xtcInput

Generic

trajectory:xtc

trrtrjgro

g96pdb

-sto

po

l.tpr

InputS

tructure+m

ass(db):tpr

tpbtpa

grog96

pdb-n

ind

ex.n

dx

Input,Opt.

Indexfile

-ol

bu

nle

n.xvg

Output

xvgr/xmgr

file-o

db

un

dist.xvg

Output

xvgr/xmgr

file-o

zb

un

z.xvgO

utputxvgr/xm

grfile

-ot

bu

ntilt.xvg

Output

xvgr/xmgr

file

Page 65: CS USER AL Simulations - GROMACS

18

4A

pp

en

dix

E.

Ma

nu

alP

age

s

E.7

gan

gle

gan

gle

com

pute

sth

ean

gle

dist

ribut

ion

for

anu

mbe

rof

angl

esor

dihe

dral

s.T

his

way

you

can

chec

kw

heth

eryo

ursi

mul

atio

nis

corr

ect.

With

optio

n-o

vyo

uca

npl

otth

eav

erag

ean

gle

ofa

grou

pof

angl

esas

afu

nctio

nof

time.

With

the

-all

optio

nth

efir

stgr

aph

isth

eav

erag

e,th

ere

star

eth

ein

divi

dual

angl

es.

With

the

-ofo

ptio

nga

ngle

also

calc

ulat

esth

efr

actio

nof

tran

sdi

hedr

als

(onl

yfo

rdi

hedr

als)

asfu

nctio

nof

time,

butt

his

ispr

obab

lyon

lyfu

nfo

ra

sele

cted

few

.

With

optio

n-o

ca

dihe

dral

corr

elat

ion

func

tion

isca

lcul

ated

.

Itsh

ould

beno

ted

that

the

inde

xfile

shou

ldco

ntai

nat

om-t

riple

sfo

rang

les

orat

om-q

uadr

uple

tsfo

rdih

edra

ls.

Ifth

isis

nott

heca

se,t

hepr

ogra

mw

illcr

ash.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Gen

eric

run

inpu

t:tp

rtp

btp

a-n

an

gle

.nd

xIn

put

Inde

xfil

e-o

da

ng

dis

t.xv

gO

utpu

txv

gr/x

mgr

file

-ov

an

ga

ver.

xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-of

dih

fra

c.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e-o

td

ihtr

an

s.xv

gO

utpu

t,O

pt.

xvgr

/xm

grfil

e-o

htr

his

to.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-o

cd

ihco

rr.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-wbo

oln

oV

iew

outp

utxv

g,xp

m,e

psan

dpd

bfil

es-t

ype

enum

an

gle

Type

ofan

gle

toan

alys

e:a

ng

le,

dih

ed

ral

,im

pro

pe

ror

ryck

ae

rt-b

elle

ma

ns

-all

bool

no

Plo

tall

angl

esse

para

tely

inth

eav

erag

esfil

e,in

the

orde

rof

appe

aran

cein

the

inde

xfil

e.-b

inw

idth

real

1bi

nwid

th(d

egre

es)

for

calc

ulat

ing

the

dist

ribut

ion

-ch

an

dle

rbo

oln

oU

seC

hand

ler

corr

elat

ion

func

tion

(N[tr

ans]

=1,

N[g

auch

e]=

0)ra

ther

than

cosi

neco

rrel

atio

nfu

nctio

n.T

rans

isde

fined

asph

i<

-60

orph

i>60

.-a

verc

orr

bool

no

Ave

rage

the

corr

elat

ion

func

tions

for

the

indi

vidu

alan

gles

/dih

edra

ls-a

cfle

nin

t-1

Leng

thof

the

AC

F,de

faul

tis

half

the

num

ber

offr

ames

-no

rma

lize

bool

yes

Nor

mal

ize

AC

F-P

enum

0O

rder

ofLe

gend

repo

lyno

mia

lfor

AC

F(0

indi

cate

sno

ne):

0,1

,2or

3-f

itfn

enum

no

ne

Fit

func

tion:

no

ne

,exp

,ae

xp,e

xpe

xpor

vac

-ncs

kip

int

0S

kip

Npo

ints

inth

eou

tput

file

ofco

rrel

atio

nfu

nctio

ns-b

eg

infit

real

0T

ime

whe

reto

begi

nth

eex

pone

ntia

lfito

fthe

corr

elat

ion

func

tion

-en

dfit

real

-1T

ime

whe

reto

end

the

expo

nent

ialfi

toft

heco

rrel

atio

nfu

nctio

n,-1

istil

lth

een

d

•C

ount

ing

tran

sitio

nson

lyw

orks

for

dihe

dral

sw

ithm

ultip

licity

3

3.1

5.

Pa

ralle

lMo

lecu

lar

Dyn

am

ics

45

Par

ticle

with

thisx

coor

dina

tera

nge

will

add

toth

ech

arge

dist

ribut

ion

onth

efo

llow

ing

rang

eof

ofF

FT

grid

slab

sin

thex

dire

ctio

n:

trun

c( il xn

x

M

) −1≤i x≤

trun

c( (i

+1)l xn

x

M

) +2

(3.8

3)

whe

retr

unc

indi

cate

sth

etr

unca

tion

ofa

real

num

ber

toth

ela

rges

tin

tege

rsm

alle

rth

anor

equa

lto

that

real

num

ber.

2.D

oing

the

Fou

rier

tran

sfor

mof

the

char

gedi

strib

utio

nρ(r

)in

para

llelt

oob

tainρ

(k).

Thi

sis

done

usin

gth

eF

FT

Wlib

rary

(seew

ww

.fftw

.org

)w

hich

empl

oys

the

MP

Ilib

rary

for

mes

sage

pass

ing

prog

ram

s(n

ote

that

ther

ear

eal

sosh

ared

mem

ory

vers

ions

ofth

eF

FT

Wco

de).

Thi

sF

FT

algo

rithm

actu

ally

use

slab

sas

wel

l(go

odth

inki

ng!)

.E

ach

proc

esso

rdo

es2D

FF

TS

onits

slab

,and

then

the

who

leF

FT

grid

istr

ansp

osed

inp

lace

(i.e.

with

outu

sing

extr

am

emor

y).

Thi

sm

eans

that

afte

rthe

FF

Tth

eX

and

Yco

mpo

nent

sar

esw

appe

d.To

com

plet

eth

eF

FT,

this

swap

ping

shou

ldbe

undo

nein

prin

cipl

e(b

ytr

ansp

osin

gba

ck).

Hap

pily

the

FF

TW

code

has

anop

tion

toom

itth

is,w

hich

we

use

inth

ene

xtst

ep.

3.C

onvo

luteρ(k

)w

ithth

eF

ourie

rtr

ansf

orm

ofth

ech

arge

spre

adfu

nctio

ng(k

)(w

hich

we

have

tabu

late

dbe

fore

)to

obta

inth

epo

tent

ial

φ(k

).A

san

optim

izat

ion,

we

stor

eth

eg(k)

intr

ansp

osed

form

asw

ell,

mat

chin

gth

etr

ansp

osed

form

ofρ(k

)w

hich

we

get

from

the

FF

TW

rout

ine.

Afte

rth

isst

epw

eha

veth

epo

tent

ial

φ(k

)in

Fou

rier

spac

e,bu

tst

illon

the

tran

spos

edF

FT

grid

.

4.D

oan

inve

rse

tran

sfor

mofφ

(k)

toob

tainφ(r

).S

ince

the

algo

rithm

mus

tdo

atr

ansp

ose

ofth

eda

tath

isst

epac

tual

lyyi

elds

the

wan

ted

resu

lt:th

eun

-tra

nspo

sed

pote

ntia

lin

real

spac

e.

5.In

terp

olat

eth

epo

tent

ialφ(r

)in

real

spac

eat

the

part

icle

posi

tions

toob

tain

forc

esan

den

ergy

.F

orth

isbi

tth

esa

me

cons

ider

atio

nsto

war

dspa

ralle

lism

hold

asfo

rth

ech

arge

spre

adin

g.H

owev

erin

this

case

mor

ene

ighb

orin

ggr

idce

llsar

ene

eded

,suc

hth

atw

ene

edth

efo

llow

ing

seto

fFF

Tgr

idsl

abs

inth

exdi

rect

ion:

trun

c( il xn

x

M

) −3≤i x≤

trun

c( (i

+1)l xn

x

M

) +4

(3.8

4)

The

algo

rithm

assk

etch

edab

ove

requ

ires

com

mun

icat

ion

fors

prea

ding

the

char

ges,

fort

heF

FT

Wfo

rwar

dan

dba

ckw

ard,

and

for

inte

rpol

atin

gth

efo

rces

.T

heG

RO

MA

CS

bits

ofth

epr

ogra

mus

eon

lyle

ftan

drig

htco

mm

unic

atio

n,i.e

.us

ing

two

com

mun

icat

ion

chan

nels

.T

heF

FT

Wro

utin

esac

tual

lyus

eot

herf

orm

sof

com

mun

icat

ion

asw

ell,

and

thes

ero

utin

esar

eco

ded

with

MP

Irou

tines

for

mes

sage

pass

ing.

Thi

sim

plie

sth

atG

RO

MA

CS

can

only

perf

orm

the

PP

PM

algo

rithm

onpa

ralle

lcom

pute

rsco

mpu

ters

that

supp

ort

MP

I.H

owev

er,

mos

tsh

ared

mem

ory

com

pute

rs,

such

asth

eS

GIO

rigin

also

supp

ortM

PIu

sing

the

shar

edm

emor

yfo

rco

mm

unic

atio

n.

3.15

.4P

aral

lels

ortin

g

For

the

dom

ain

deco

mpo

sitio

nbi

tofG

RO

MA

CS

itis

nece

ssar

yto

sort

the

coor

dina

tes

(or

rath

erth

ein

dex

toco

ordi

nate

s)ev

ery

time

ane

ighb

orlis

tis

mad

e.If

we

use

brut

efo

rce,

and

sort

all

Page 66: CS USER AL Simulations - GROMACS

46

Ch

ap

ter

3.

Alg

orith

ms

coordinateson

eachprocessor

(which

istechnically

possiblesince

we

haveallthe

coordinates),then

thissorting

procedurew

illtakea

constanttime

(proportionaltoN

2logN

,independentofthenum

berof

processors.W

ecan

however

doa

littlebetter,

ifw

eassum

ethat

particlesdiffuse

onlyslow

ly.A

parallelsortingalgorithm

canbe

conceivedas

follows:

Atthe

firststepofthe

simulation

1.D

oa

fullsortofallindicesusing

e.g.thequick-sortalgorithm

thatisbuilt-in

inthe

standardC

-library

2.D

ividethe

sortedarray

intoslabs

(asdescribed

abovesee

Fig.

3.14).

Atsubsequentsteps

ofthesim

ulation:

1.S

endthe

indicesfor

eachprocessor

tothe

precedingprocessor

(ifnot

processor0)

andto

thenext

processor(if

notM-1).

The

comm

unicationassociated

with

thisoperation

isproportionalto

2N/M

.

2.S

ortthecom

binedindices

ofthethree

(ortw

o)processors.

Note

thattheC

PU

time

associ-ated

with

sortingis

now(3N

/M) 2log

(3N/M

).

3.O

neach

processor,the

indicesbelonging

toit’s

slabcan

bedeterm

inedfrom

theorder

ofthe

array(F

ig.3.14).

E.6

.g

an

alyze

18

3

This

isusefulfor

principalcomponents

obtainedfrom

covarianceanalysis,since

theprincipalcom

ponentsofrandom

diffusionare

purecosines.

Option

-msd

producesthe

mean

squaredisplacem

ent(s).

Option

-dist

producesdistribution

plot(s).

Option

-av

producesthe

averageover

thesets.

Error

barscan

beadded

with

theoption

-errb

ar

.T

heerrorbars

canrepresentthe

standarddeviation,the

error(assuming

thepoints

areindependent)orthe

intervalcontaining

90%ofthe

points,bydiscarding

5%ofthe

pointsatthe

topand

thebottom

.

Option

-ee

produceserror

estimates

usingblock

averaging.A

setis

dividedin

anum

berof

blocksand

averagesare

calculatedfor

eachblock.

The

errorfor

thetotal

averageis

calculatedfrom

thevariance

between

averagesofthe

mblocks

Biasfollow

s:error 2

=S

um(B

i-<

B>

) 2/(m

*(m-1)).

These

errorsare

plottedas

afunction

oftheblock

size.A

lsoan

analyticalblockaverage

curveis

plotted,assuming

thattheautocorrelation

isa

sumoftw

oexponentials.

The

analyticalcurvefor

theblock

averageB

Ais:

BA

(t)=

sigma

sqrt(2/T(

a(tau1

((exp(-t/tau1)-

1)tau1/t+

1))+

(1-a)(tau2

((exp(-t/tau2)-

1)tau2/t+

1)))),w

hereT

isthe

totaltime.

a,tau1

andtau2

areobtained

byfitting

BA

(t)to

thecalculated

blockaverage.

When

theactualblock

averageis

veryclose

tothe

analyticalcurve,the

erroris

sigma*sqrt(2/T

(atau1

+(1-a)

tau2)).

Option

-po

we

rfits

thedata

tob

ta,which

isaccom

plishedby

fittingto

at+

bon

log-logscale.

Allpoints

afterthe

firstzeroor

negativevalue

areignored.

Files

-fg

rap

h.xvg

Inputxvgr/xm

grfile

-ac

au

toco

rr.xvgO

utput,Opt.

xvgr/xmgr

file-m

sdm

sd.xvg

Output,O

pt.xvgr/xm

grfile

-ccco

scon

t.xvgO

utput,Opt.

xvgr/xmgr

file-d

istd

istr.xvgO

utput,Opt.

xvgr/xmgr

file-a

va

vera

ge

.xvgO

utput,Opt.

xvgr/xmgr

file-e

ee

rrest.xvg

Output,O

pt.xvgr/xm

grfile

Other

options-h

booln

oP

rinthelpinfo

andquit

-nice

int1

9S

etthenicelevel

-wbool

no

View

outputxvg,xpm,eps

andpdb

files-tim

ebool

yes

Expecta

time

inthe

input-b

real-1

Firsttim

eto

readfrom

set-e

real-1

Lasttime

toread

fromset

-nint

1R

ead#

setsseperated

by&

-dbool

no

Use

thederivative

-bw

real0

.1B

inwidth

forthe

distribution-e

rrba

renum

no

ne

Error

barsfor

-av:no

ne

,stdd

ev

,erro

ror9

0-p

ow

er

booln

oF

itdatato:

bta

-sub

av

boolye

sS

ubtracttheaverage

beforeautocorrelating

-on

ea

cfbool

no

Calculate

oneA

CF

overallsets

-acfle

nint

-1Length

oftheA

CF,defaultis

halfthenum

beroffram

es-n

orm

alize

boolye

sN

ormalize

AC

F-P

enum0

Order

ofLegendrepolynom

ialforA

CF

(0indicates

none):0

,1,2

or3-fitfn

enumn

on

eF

itfunction:no

ne

,exp

,ae

xp,e

xpe

xporva

c-n

cskipint

0S

kipN

pointsin

theoutputfile

ofcorrelationfunctions

-be

gin

fitreal

0T

ime

where

tobegin

theexponentialfitofthe

correlationfunction

-en

dfit

real-1

Tim

ew

hereto

endthe

exponentialfitofthecorrelation

function,-1is

tillthe

end

Page 67: CS USER AL Simulations - GROMACS

18

2A

pp

en

dix

E.

Ma

nu

alP

age

s

whe

reM

1an

dM

2ar

eth

etw

oco

varia

nce

mat

rices

and

tris

the

trac

eof

am

atrix

.T

henu

mbe

rsar

epr

o-po

rtio

nalt

oth

eov

erla

pof

the

squa

rero

otof

the

fluct

uatio

ns.

The

norm

aliz

edov

erla

pis

the

mos

tus

eful

num

ber,

itis

1fo

rid

entic

alm

atric

esan

d0

whe

nth

esa

mpl

edsu

bspa

ces

are

orth

ogon

al.

File

s-v

eig

en

vec.

trr

Inpu

tF

ullp

reci

sion

traj

ecto

ry:

trr

trj

-v2

eig

en

vec2

.trr

Inpu

t,O

pt.

Ful

lpre

cisi

ontr

ajec

tory

:tr

rtr

j-f

tra

j.xtc

Inpu

t,O

pt.

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b-s

top

ol.t

pr

Inpu

t,O

pt.

Str

uctu

re+

mas

s(db

):tp

rtp

btp

agr

og9

6pd

b-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-eig

1e

ige

nva

l1.x

vgIn

put,

Opt

.xv

gr/x

mgr

file

-eig

2e

ige

nva

l2.x

vgIn

put,

Opt

.xv

gr/x

mgr

file

-dis

pe

igd

isp

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

-pro

jp

roj.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-2

d2

dp

roj.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-3

d3

dp

roj.p

db

Out

put,

Opt

.G

ener

icst

ruct

ure:

gro

g96

pdb

-filt

filte

red

.xtc

Out

put,

Opt

.G

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-ext

re

xtre

me

.pd

bO

utpu

t,O

pt.

Gen

eric

traj

ecto

ry:

xtc

trr

trjg

rog9

6pd

b-o

ver

ove

rla

p.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-in

pr

inp

rod

.xp

mO

utpu

t,O

pt.

XP

ixM

apco

mpa

tible

mat

rixfil

e

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-tu

enum

ps

Tim

eun

it:p

s,f

s,n

s,u

s,m

s,s

,mor

h-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-first

int

1F

irste

igen

vect

orfo

ran

alys

is(-

1is

sele

ct)

-la

stin

t8

Last

eige

nvec

tor

for

anal

ysis

(-1

istil

lthe

last

)-s

kip

int

1O

nly

anal

yse

ever

ynr

-th

fram

e-m

ax

real

0M

axim

umfo

rpr

ojec

tion

ofth

eei

genv

ecto

ron

the

aver

age

stru

ctur

e,m

ax=

0gi

ves

the

extr

emes

-nfr

am

es

int

2N

umbe

rof

fram

esfo

rth

eex

trem

esou

tput

-sp

litbo

oln

oS

plit

eige

nvec

tor

proj

ectio

nsw

here

time

isze

ro

E.6

gan

alyz

e

gan

alyz

ere

ads

anas

ciifi

lean

dan

alyz

esda

tase

ts.

Alin

ein

the

inpu

tfile

may

star

twith

atim

e(s

eeop

tion

-tim

e)

and

any

num

ber

ofy

valu

esm

ayfo

llow

.M

ultip

lese

tsca

nal

sobe

read

whe

nth

eyar

ese

pera

ted

by&

(opt

ion

-n),

inth

isca

seon

lyon

ey

valu

eis

read

from

each

line.

All

lines

star

ting

with

#an

d@

are

skip

ped.

All

anal

yses

can

also

bedo

nefo

rth

ede

rivat

ive

ofa

set(

optio

n-d

).

All

optio

ns,e

xcep

tfor

-av

and

-po

we

ras

sum

eth

atth

epo

ints

are

equi

dist

anti

ntim

e.

gan

alyz

eal

way

ssh

ows

the

aver

age

and

stan

dard

devi

atio

nof

each

set.

For

each

set

ital

sosh

ows

the

rela

tive

devi

atio

nof

the

third

and

fort

hcu

mul

ant

from

thos

eof

aG

auss

ian

dist

ribut

ion

with

the

sam

est

anda

rdde

viat

ion.

Opt

ion

-ac

prod

uces

the

auto

corr

elat

ion

func

tion(

s).

Opt

ion

-cc

plot

sth

ere

sem

blan

ceof

seti

with

aco

sine

ofi/2

perio

ds.

The

form

ula

is:

2(in

t0-T

y(t)

cos(

pit/i

)dt

)2/i

nt0-

Ty(

t)y(

t)dt

Cha

pter

4

For

cefie

lds

Afo

rce

field

isbu

iltup

from

two

dist

inct

com

pone

nts:

•T

hese

tofe

quat

ions

(cal

led

thepo

ten

tialf

un

ctio

ns)us

edto

gene

rate

the

pote

ntia

lene

rgie

san

dth

eir

deriv

ativ

es,t

hefo

rces

.

•T

hepa

ram

eter

sus

edin

this

seto

fequ

atio

ns

With

inon

ese

tof

equa

tions

vario

usse

tsof

para

met

ers

can

beus

ed.

Car

em

ust

beta

ken

that

the

com

bina

tion

ofeq

uatio

nsan

dpa

ram

eter

sfo

rma

cons

iste

ntse

t.It

isin

gene

rald

ange

rous

tom

ake

ad

ho

ccha

nges

ina

subs

etof

para

met

ers,

beca

use

the

vario

usco

ntrib

utio

nsto

the

tota

lfor

cear

eus

ually

inte

rdep

ende

nt.

InG

RO

MA

CS

3.0

the

forc

efie

ldis

base

don

GR

OM

OS

-87

[39

],w

itha

smal

lmod

ifica

tion

con-

cern

ing

the

inte

ract

ion

betw

een

wat

er-o

xyge

nsan

dca

rbon

atom

s[

40,4

1],a

sw

ella

s10

extr

aat

omty

pes

[42,

43,4

0,41

,44]

.H

owev

er,t

heus

eris

free

tom

ake

hero

wn

mod

ifica

tions

(bew

are!

).T

his

will

beex

plai

ned

inde

tails

inch

apte

r5,w

hich

deal

sw

ithth

eTop

olog

y.To

acco

mm

odat

eth

epo

tent

ial

func

tions

used

inso

me

popu

lar

forc

efie

lds,

GR

OM

AC

Sof

fers

ach

oice

offu

nctio

ns,

both

for

non-

bond

edin

tera

ctio

nan

dfo

rdi

hedr

alin

tera

ctio

ns.

The

yar

ede

scrib

edin

the

appr

opria

tesu

bsec

tions

.

The

pote

ntia

lfun

ctio

nsca

nbe

subd

ivid

edin

toth

ree

part

s

1.N

on

-bo

nd

ed:

Lenn

ard-

Jone

sor

Buc

king

ham

,and

Cou

lom

bor

mod

ified

Cou

lom

b.T

heno

n-bo

nded

inte

ract

ions

are

com

pute

don

the

basi

sof

ane

ighb

orlis

t(a

listo

fnon

-bon

ded

atom

sw

ithin

ace

rtai

nra

dius

),in

whi

chex

clus

ions

are

alre

ady

rem

oved

.

2.B

on

de

d:co

vale

ntbo

nd-s

tret

chin

g,an

gle-

bend

ing,

impr

oper

dihe

dral

s,an

dpr

oper

dihe

dral

s.T

hese

are

com

pute

don

the

basi

sof

fixed

lists

.

3.S

pe

cia

l:po

sitio

nre

stra

ints

and

dist

ance

rest

rain

ts,b

ased

onfix

edlis

ts.

Page 68: CS USER AL Simulations - GROMACS

48

Ch

ap

ter

4.

Force

field

s

0.40.5

0.60.7

0.8r (nm

)

–0.2

0.0

0.2

0.4

V (kJ mole–1

)Figure

4.1:T

heLennard-Jones

interaction.

4.1N

on-bondedinteractions

Non-bonded

interactionsin

GR

OM

AC

Sare

pair-additiveand

centro-symm

etric:

V(r

1 ,...rN

)= ∑i<

j

Vij (r

ij );(4.1)

Fi =

− ∑j

dV

ij (rij )

drij

rij

rij

=−

Fj

(4.2)

The

non-bondedinteractions

containa

repulsionterm

,a

dispersionterm

,and

aC

oulomb

term.

The

repulsionand

dispersionterm

arecom

binedin

eitherthe

Lennard-Jones(or

6-12interaction),

orthe

Buckingham

(orexp-6

potential).In

addition,(partially)

chargedatom

sact

throughthe

Coulom

bterm

.

4.1.1T

heLennard-Jones

interaction

The

Lennard-Jonespotential

VL

Jbetw

eentw

oatom

sequals

VL

J (rij )

=C

(12)

ij

r12

ij

−C

(6)

ij

r6ij

(4.3)

seealso

Fig.4.1

The

parametersC

(12)

ijand

C(6

)ij

dependon

pairsofato

mtyp

es;consequently

theyare

takenfrom

am

atrixofLJ-param

eters.

The

forcederived

fromthis

potentialis:

Fi (r

ij )=

12C

(12)

ij

r12

ij

−6C

(6)

ij

r6ij

rij

rij

(4.4)

E.5

.g

an

ae

ig1

81

-nice

int1

9S

etthenicelevel

-breal

-1F

irsttime

touse

-ereal

-1Lasttim

eto

use-d

treal

0O

nlyw

riteoutfram

ew

hentM

OD

dt=offset

-offse

treal

0T

ime

offsetfor-dtoption

-settim

ebool

no

Change

startingtim

einteractively

-sort

boolye

sS

ortenergyfiles

(notframes)

-scale

fac

real1

Multiply

energycom

ponentbythis

factor-e

rror

boolye

sS

topon

errorsin

thefile

•W

hencom

biningtrajectories

thesigm

aand

E2

(necessaryfor

statistics)are

notupdated

correctly.O

nlythe

actualenergyis

correct.O

nethus

hasto

compute

statisticsin

anotherw

ay.

E.5

ganaeig

ga

na

eig

analyzeseigenvectors.

The

eigenvectorscan

beof

acovariance

matrix

(g

cova

r)

orof

aN

ormalM

odesanaysis

(g

nm

eig

).

When

atrajectory

isprojected

oneigenvectors,

allstructuresare

fittedto

thestructure

inthe

eigenvectorfile,ifpresent,otherw

iseto

thestructure

inthe

structurefile.

When

norun

inputfileis

supplied,periodicityw

illnotbetaken

intoaccount.

Mostanalyses

areperform

edon

eigenvectors-first

to-la

st,butw

hen-first

issetto

-1you

willbe

prompted

fora

selection.

-disp

:plotallatom

displacements

ofeigenvectors-first

to-la

st.

-pro

j:

calculateprojections

ofa

trajectoryon

eigenvectors-first

to-la

st.

The

projectionsof

atrajectory

onthe

eigenvectorsof

itscovariance

matrix

arecalled

principalcomponents

(pc’s).It

isoften

usefultocheck

thecosine

contentthepc’s,since

thepc’s

ofrandomdiffusion

arecosines

with

thenum

berof

periodsequalto

halfthe

pcindex.

The

cosinecontent

ofthe

pc’scan

becalculated

with

theprogram

ga

na

lyze.

-2d

:calculate

a2d

projectionofa

trajectoryon

eigenvectors-first

and-la

st.

-3d

:calculate

a3d

projectionofa

trajectoryon

thefirstthree

selectedeigenvectors.

-filt:

filterthe

trajectoryto

showonly

them

otionalong

eigenvectors-first

to-la

st.

-extr

:calculate

thetw

oextrem

eprojections

alonga

trajectoryon

theaverage

structureand

interpolate-n

fram

es

frames

between

them,or

setyourow

nextrem

esw

ith-m

ax

.T

heeigenvector-first

willbe

written

unless-firstand

-last

havebeen

setexplicitly,inw

hichcase

alleigenvectorsw

illbew

rittento

separatefiles.

Chain

identifiersw

illbeadded

when

writing

a.p

db

filew

ithtw

oor

threestructures

(youcan

userasm

ol

-nm

rpd

bto

viewsuch

apdb

file).

Overlap

calculationsbetw

eencovariance

analysis:N

OT

E:the

analysisshould

usethe

same

fittingstructure

-ove

r:

calculatethe

subspaceoverlap

oftheeigenvectors

infile

-v2w

itheigenvectors-first

to-la

stin

file-v

.

-inp

r:

calculatea

matrix

ofinner-products

between

eigenvectorsin

files-v

and-v2

.A

lleigenvectorsofboth

filesw

illbeused

unless-firstand

-last

havebeen

setexplicitly.

When

-v,

-eig

1,

-v2and

-eig

2are

given,a

singlenum

berfor

theoverlap

between

thecovariance

matrices

isgenerated.

The

formulas

are:difference

=sqrt(tr((sqrt(M

1)-

sqrt(M2)) 2))

normalized

overlap=

1-

difference/sqrt(tr(M1)

+tr(M

2))shape

overlap=

1-

sqrt(tr((sqrt(M1/tr(M

1))-

sqrt(M2/tr(M

2)))2))

Page 69: CS USER AL Simulations - GROMACS

18

0A

pp

en

dix

E.

Ma

nu

alP

age

s

-fco

nf.g

roIn

put

Gen

eric

stru

ctur

e:gr

og9

6pd

btp

rtp

btp

a-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-oo

ut.g

roO

utpu

tG

ener

icst

ruct

ure:

gro

g96

pdb

-bf

bfa

ct.d

at

Inpu

t,O

pt.

Gen

eric

data

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t0

Set

the

nice

leve

l-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-nd

ef

bool

no

Cho

ose

outp

utfr

omde

faul

tind

exgr

oups

-bt

enum

tric

Box

type

for

-box

and

-d:

tric

,cu

bic

,d

od

eca

he

dro

nor

oct

ah

ed

ron

-bo

xve

ctor

00

0B

oxve

ctor

leng

ths

(a,b

,c)

-an

gle

sve

ctor9

09

09

0A

ngle

sbe

twee

nth

ebo

xve

ctor

s(b

c,ac

,ab)

-dre

al0

Dis

tanc

ebe

twee

nth

eso

lute

and

the

box

-cbo

oln

oC

ente

rm

olec

ule

inbo

x(im

plie

dby

-box

and

-d)

-ce

nte

rve

ctor

00

0C

oord

inat

esof

geom

etric

alce

nter

-ro

tate

vect

or0

00

Rot

atio

nar

ound

the

X,Y

and

Zax

esin

degr

ees

-prin

cbo

oln

oO

rient

mol

ecul

e(s)

alon

gth

eir

prin

cipa

laxe

s-s

cale

vect

or1

11

Sca

ling

fact

or-d

en

sity

real

10

00

Den

sity

(g/l)

ofth

eou

tput

box

achi

eved

bysc

alin

g-p

bc

bool

no

Rem

ove

the

perio

dici

ty(m

ake

mol

ecul

ew

hole

agai

n)-m

ea

dbo

oln

oS

tore

the

char

geof

the

atom

inth

eoc

cupa

ncy

field

and

the

radi

usof

the

atom

inth

eB

-fac

tor

field

-gra

spbo

oln

oS

tore

the

char

geof

the

atom

inth

eB

-fac

tor

field

and

the

radi

usof

the

atom

inth

eoc

cupa

ncy

field

-rvd

wre

al0

.12

Def

ault

Van

der

Waa

lsra

dius

ifon

eca

nno

tbe

foun

din

the

data

base

-ato

mbo

oln

oF

orce

B-f

acto

rat

tach

men

tper

atom

-le

ge

nd

bool

no

Mak

eB

-fac

tor

lege

nd-la

be

lst

ring

AA

ddch

ain

labe

lfor

allr

esid

ues

•F

orco

mpl

exm

olec

ules

,th

epe

riodi

city

rem

oval

rout

ine

may

brea

kdo

wn,

inth

atca

seyo

uca

nus

etr

jcon

v

E.4

enec

onv

Whe

n-f

isn

ots

peci

fied:

Con

cate

nate

sse

vera

lene

rgy

files

inso

rted

orde

r.In

case

ofdo

uble

time

fram

esth

eon

ein

the

late

rfil

eis

used

.B

ysp

ecify

ing-s

ettim

eyo

uw

illbe

aske

dfo

rth

est

artt

ime

ofea

chfil

e.T

hein

putfi

les

are

take

nfr

omth

eco

mm

and

line,

such

that

the

com

man

de

ne

con

v-o

fixe

d.e

dr

*.e

dr

shou

lddo

the

tric

k.

With

-fsp

ecifi

ed:

Rea

dson

een

ergy

file

and

writ

esan

othe

r,ap

plyi

ngth

e-d

t,-

offse

t,-

t0an

d-s

ettim

eop

tions

and

conv

ertin

gto

adi

ffere

ntfo

rmat

ifne

cess

ary

(indi

cate

dby

file

exte

ntio

ns).

-se

ttim

eis

appl

ied

first

,the

n-dt

/-o

ffse

tfo

llow

edby

-ban

d-e

tose

lect

whi

chfr

ames

tow

rite.

File

s-f

en

er.

ed

rIn

put

Gen

eric

ener

gy:

edr

ene

-ofix

ed

.ed

rO

utpu

t,O

pt.

Gen

eric

ener

gy:

edr

ene

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

4.1

.N

on

-bo

nd

ed

inte

ract

ion

s4

9

0.2

0.3

0.4

0.5

0.6

0.7

0.8

r (n

m)

–0.5

0.0

0.5

1.0

1.5

V (kJ mole–1

) Fig

ure

4.2:

The

Buc

king

ham

inte

ract

ion.

The

LJpo

tent

ialm

ayal

sobe

writ

ten

inth

efo

llow

ing

form

:

VL

J(r

ij)

=4ε

ij

( σ ij r ij

) 12−( σ ij r i

j

) 6 (4

.5)

Inco

nstr

uctin

gth

epa

ram

eter

mat

rixfo

rth

eno

n-bo

nded

LJ-p

aram

eter

s,tw

oty

pes

ofco

mbi

natio

nru

les

can

beus

edw

ithin

GR

OM

AC

S:

C(6

)ij

=( C

(6)

ii∗C

(6)

jj

) 1/2C

(12)

ij=

( C(1

2)

ii∗C

(12)

jj

) 1/2(4

.6)

or,a

ltern

ativ

ely,

σij

=1 2(σ

ii+σ

jj)

ε ij

=(ε

iiε j

j)1

/2

(4.7

)

4.1.

2B

ucki

ngha

mpo

tent

ial

The

Buc

king

ham

pote

ntia

lhas

am

ore

flexi

ble

and

real

istic

repu

lsio

nte

rmth

anth

eLe

nnar

d-Jo

nes

inte

ract

ion,

buti

sal

som

ore

expe

nsiv

eto

com

pute

.T

hepo

tent

ialf

orm

is:

Vbh

(rij)

=A

ijex

p(−B

ijr i

j)−C

ij

r6 ij

(4.8

)

see

also

Fig

.4.2,

the

forc

ede

rived

from

this

is:

Fi(r i

j)

=

[ −A

ijB

ijr i

jex

p(−B

ijr i

j)−

6C

ij

r6 ij

] rij

r ij

(4.9

)

Page 70: CS USER AL Simulations - GROMACS

50

Ch

ap

ter

4.

Force

field

s

0.00.2

0.40.6

0.81.0

r (nm)

0

500

1000

1500

V (kJ mol� −1)

Coulom

bW

ith RF

�RF

− C

Figure

4.3:T

heC

oulomb

interaction(for

particlesw

ithequal

signedcharge)

with

andw

ithoutreaction

field.In

thelatter

caseεrf

was

78,andrc

was

0.9nm

.T

hedot-dashed

lineis

thesam

eas

thedashed

line,exceptfora

constant.

4.1.3C

oulomb

interaction

The

Coulom

binteraction

between

two

chargeparticles

isgiven

by:

Vc (r

ij )=fqi q

j

εr r

ij(4.10)

seealso

Fig.4.3,w

heref=

14πε0

=138.935

485(see

chapter2)

The

forcederived

fromthis

potentialis:

Fi (r

ij )=fqi q

j

εr r

2ij

rij

rij

(4.11)

InG

RO

MA

CS

therelative

dielectricconstant

εr

may

besetin

thein

theinputforgro

mp

p.

4.1.4C

oulomb

interactionw

ithreaction

field

The

coulomb

interactioncan

bem

odifiedfor

homogeneous

systems,

byassum

inga

constantdi-

electricenvironm

entbeyond

thecut-off

rc

with

adielectric

constantofεr

f .T

heinteraction

thenreads:

Vcr

f=

fqi q

j

rij [1

+εrf−

12ε

rf

+1r3ij

r3c ]

−fqi q

j

rc

3εrf

2εrf

+1

(4.12)

inw

hichthe

constantexpression

onthe

rightm

akesthe

potentialzeroat

thecut-offrc .

We

canrew

ritethis

forsim

plicityas

Vcr

f=

fqi q

j [1rij

+k

rfr2ij −

crf ]

(4.13)

E.3

.e

ditco

nf

17

9

E.3

editconf

editconfconvertsgeneric

structureform

atto.g

ro,.g

96

or.pd

b.

The

boxcan

bem

odifiedw

ithoptions-bo

x,-d

and-a

ng

les

.B

oth-b

ox

and-d

willcenter

thesystem

inthe

box.

Option

-bt

determines

thebox

type:tricis

atriclinic

box,cub

icis

acubic

box,do

de

cah

ed

ron

isa

rhombic

dodecahedronandocta

he

dro

nis

atruncated

octahedron.T

helast

two

arespecialcases

ofa

triclinicbox.

The

lengthof

thethree

boxvectors

ofthe

truncatedoctahedron

isthe

shortestdistance

between

two

oppositehexagons.

The

volume

ofadodecahedron

is0.71

andthatofa

truncatedoctahedron

is0.77

ofthatofacubic

boxw

iththe

same

periodicim

agedistance.

Option

-bo

xrequires

onlyone

valuefor

acubic

box,dodecahedronand

atruncated

octahedron.W

ith-d

andtric

thesize

ofthesystem

inthe

x,yand

zdirections

isused.

With

-dand

cub

ic,d

od

eca

he

dro

noro

ctah

ed

ron

thediam

eterofthe

systemis

used,which

isthe

largestdistancebetw

eentw

oatom

s.

Option

-an

gle

sis

onlym

eaningfulwith

option-bo

xand

atriclinic

boxand

cannotbe

usedw

ithoption

-d.

When

-nor

-nd

ef

isset,a

groupcan

beselected

forcalculating

thesize

andthe

geometric

center,other-w

isethe

whole

systemis

used.

-rota

terotates

thecoordinates

andvelocities.

-prin

caligns

theprincipalaxes

ofthe

systemalong

thecoordinate

axes,this

may

allowyou

todecrease

thebox

volume,

butbew

arethat

molecules

canrotate

significantlyin

ananosecond.

Scaling

isapplied

beforeany

ofthe

otheroperations

areperform

ed.B

oxescan

bescaled

togive

acertain

density(option-d

en

sity).

Aspecialfeature

ofthe

scalingoption,

when

thefactor

-1is

givenin

onedim

ension,oneobtains

am

irrorim

age,mirrored

inone

oftheplains,w

henone

uses-1

inthree

dimensions

apoint-m

irrorim

ageis

obtained.

Groups

areselected

afteralloperations

havebeen

applied.

Periodicity

canbe

removed

ina

crudem

anner.Itis

importantthatthe

boxsizes

atthebottom

ofyourinput

fileare

correctwhen

theperiodicity

isto

berem

oved.

The

programcan

optionallyrotate

thesolute

molecule

toalign

them

oleculealong

itsprincipal

axes(-ro

tate

)

When

writing

.pd

bfiles,B

-factorscan

beadded

with

the-bfoption.

B-factors

areread

froma

filew

ithw

ithfollow

ingform

at:firstline

statesnum

berofentries

inthe

file,nextlinesstate

anindex

followed

bya

B-factor.

The

B-factors

willbe

attachedper

residueunless

anindex

islarger

thanthe

number

ofresidues

orunless

the-ato

moption

isset.

Obviously,any

typeofnum

ericdata

canbe

addedinstead

ofB-factors.

-leg

en

dw

illproduce

arow

ofC

Aatom

sw

ithB

-factorsranging

fromthe

minim

umto

them

aximum

valuefound,effectively

making

alegend

forview

ing.

With

theoption

-mead

aspecialpdb

filefor

theM

EA

Delectrostatics

program(P

oisson-Boltzm

annsolver)

canbe

made.

Afurther

prerequisiteis

thattheinputfile

isa

runinputfile.

The

B-factor

fieldis

thenfilled

with

theVan

derW

aalsradius

oftheatom

sw

hilethe

occupancyfield

willhold

thecharge.

The

option-grasp

issim

ilar,butitputsthe

chargesin

theB

-factorand

theradius

inthe

occupancy.

Finally

with

option-la

be

leditconf

canadd

achain

identifierto

apdb

file,w

hichcan

beuseful

foranalysis

with

e.g.rasm

ol.

Toconverta

truncatedoctrahedron

fileproduced

bya

packagew

hichuses

acubic

boxw

iththe

cornerscut

off(suchas

Grom

os)use:

ed

itcon

f-f

<in>

-rota

te0

-45

-35

.26

4-b

to

-bo

x<

vecle

n>

-o<

ou

t>

whereve

clen

isthe

sizeofthe

cubicbox

times

sqrt(3)/2.F

iles

Page 71: CS USER AL Simulations - GROMACS

17

8A

pp

en

dix

E.

Ma

nu

alP

age

s

•M

ost

GR

OM

AC

Spr

ogra

ms

can

proc

ess

atr

ajec

tory

with

less

atom

sth

anth

eru

nin

put

orst

ruct

ure

file,

buto

nly

ifth

etr

ajec

tory

cons

ists

ofth

efir

stn

atom

sof

the

run

inpu

tor

stru

ctur

efil

e.

•M

any

GR

OM

AC

Spr

ogra

ms

will

acce

ptth

e-tu

optio

nto

set

the

time

units

tous

ein

outp

utfil

es(e

.g.

forx

mg

rgr

aphs

orxp

mm

atric

es)

and

inal

ltim

eop

tions

.

E.2

dods

sp

dods

spre

ads

atr

ajec

tory

file

and

com

pute

sth

ese

cond

ary

stru

ctur

efo

rea

chtim

efr

ame

calli

ngth

eds

sppr

ogra

m.

Ifyo

udo

not

have

the

dssp

prog

ram

,ge

tit.

dodssp

assu

mes

that

the

dssp

exec

utab

leis

in/h

ome/

mdg

roup

/dss

p/ds

sp.

Ifth

atis

nott

heca

se,t

hen

you

shou

ldse

tan

envi

ronm

entv

aria

ble

DS

SP

poin

t-in

gto

the

dssp

exec

utab

leas

in:

sete

nv

DS

SP

/usr

/loca

l/bin

/dss

p

The

stru

ctur

eas

sign

men

tfo

rea

chre

sidu

ean

dtim

eis

writ

ten

toan

.xp

mm

atrix

file.

Thi

sfil

eca

nbe

visu

aliz

edw

ithfo

rin

stan

cexv

and

can

beco

nver

ted

topo

stsc

riptw

ithxpm

2p

s.

The

num

ber

ofre

sidu

esw

ithea

chse

cond

ary

stru

ctur

ety

pean

dth

eto

tals

econ

dary

stru

ctur

e(

-sss

)co

unta

sa

func

tion

oftim

ear

eal

sow

ritte

nto

file

(-sc

).

Sol

vent

acce

ssib

lesu

rfac

e(S

AS

)per

resi

due

can

beca

lcul

ated

,bot

hin

abso

lute

valu

es(A

2)a

ndin

frac

tions

ofth

em

axim

alac

cess

ible

surf

ace

ofa

resi

due.

The

max

imal

acce

ssib

lesu

rfac

eis

defin

edas

the

acce

ssib

lesu

rfac

eof

are

sidu

ein

ach

ain

ofgl

ycin

es.

Not

eth

atth

epr

ogra

mgsa

sca

nal

soco

mpu

teS

AS

and

that

ism

ore

effic

ient

.

Fin

ally

,th

ispr

ogra

mca

ndu

mp

the

seco

ndar

yst

ruct

ure

ina

spec

ial

file

ssd

um

p.d

at

for

usag

ein

the

prog

ram

gch

i.

Toge

ther

thes

etw

opr

ogra

ms

can

beus

edto

anal

yze

dihe

dral

prop

ertie

sas

afu

nctio

nof

seco

ndar

yst

ruct

ure

type

.

File

s-f

tra

j.xtc

Inpu

tG

ener

ictr

ajec

tory

:xt

ctr

rtr

jgro

g96

pdb

-sto

po

l.tp

rIn

put

Str

uctu

re+

mas

s(db

):tp

rtp

btp

agr

og9

6pd

b-n

ind

ex.

nd

xIn

put,

Opt

.In

dex

file

-ssd

um

pss

du

mp

.da

tO

utpu

t,O

pt.

Gen

eric

data

file

-ma

pss

.ma

pIn

put,

Lib.

File

that

map

sm

atrix

data

toco

lors

-oss

.xp

mO

utpu

tX

Pix

Map

com

patib

lem

atrix

file

-sc

sco

un

t.xv

gO

utpu

txv

gr/x

mgr

file

-aa

rea

.xp

mO

utpu

t,O

pt.

XP

ixM

apco

mpa

tible

mat

rixfil

e-t

ato

tare

a.x

vgO

utpu

t,O

pt.

xvgr

/xm

grfil

e-a

aa

vera

rea

.xvg

Out

put,

Opt

.xv

gr/x

mgr

file

Oth

erop

tions

-hbo

oln

oP

rinth

elp

info

and

quit

-nic

ein

t1

9S

etth

eni

cele

vel

-btim

e-1

Firs

tfra

me

(ps)

tore

adfr

omtr

ajec

tory

-etim

e-1

Last

fram

e(p

s)to

read

from

traj

ecto

ry-d

ttim

e-1

Onl

yus

efr

ame

whe

ntM

OD

dt=

first

time

(ps)

-tu

enum

ps

Tim

eun

it:p

s,f

s,n

s,u

s,m

s,s

,mor

h-w

bool

no

Vie

wou

tput

xvg,

xpm

,eps

and

pdb

files

-sss

strin

gH

EB

TS

econ

dary

stru

ctur

esfo

rst

ruct

ure

coun

t

•T

hepr

ogra

mis

very

slow

4.1

.N

on

-bo

nd

ed

inte

ract

ion

s5

1

with

krf

=1 r3 c

ε rf−

1(2ε r

f+

1)(4

.14)

c rf

=1 r c

+k

rfr2 c

=1 r c

3εrf

(2ε r

f+

1)(4

.15)

for

larg

eε r

fth

ek

rf

goes

to0.

5r−

3c

,w

hile

forε r

f=

1th

eco

rrec

tion

vani

shes

.T

his

mak

esit

poss

ible

tous

eth

esa

me

expr

essi

onw

ithan

dw

ithou

trea

ctio

nfie

ld,a

lbei

tats

ome

com

puta

tiona

lco

st.

InF

ig.4

.3th

em

odifi

edin

tera

ctio

nis

plot

ted,

and

itis

clea

rth

atth

ede

rivat

ive

with

resp

ect

tor i

j(=

-for

ce)

goes

toze

roat

the

cut-

offd

ista

nce.

The

forc

ede

rived

from

this

pote

ntia

lrea

ds:

Fi(

rij)

=fq iq j

[ 1 r2 ij

−2k

rfr i

j

] rij

r ij

(4.1

6)

Tiro

nie

tal.

have

intr

oduc

eda

gene

raliz

edre

actio

nfie

ldin

whi

chth

edi

elec

tric

cont

inuu

mbe

yond

the

cut-

offr

cal

soha

san

ioni

cst

reng

thI[4

5].

Inth

isca

sew

eca

nre

writ

eth

eco

nsta

nts

krf

and

c rf

usin

gth

ein

vers

eD

ebye

scre

enin

gle

ngth

κ:

κ=

2IF

2

ε 0ε r

fRT

=F

2

ε 0ε r

fRT

K ∑ i=1

c iz i

(4.1

7)

krf

=1 r3 c

(εrf−

1)(1

+κr c

)+ε r

f(κr c

)2

(2ε r

f+

1)(1

+κr c

)+

2εrf(κr c

)2(4

.18)

c rf

=1 r c

3εrf(1

+κr c

+(κr c

)2)

(2ε r

f+

1)(1

+κr c

)+

2εrf(κr c

)2(4

.19)

whe

reF

isF

arad

ay’s

cons

tant

,Ris

the

idea

lga

sco

nsta

nt,T

the

abso

lute

tem

pera

ture

,c i

the

mol

arco

ncen

trat

ion

fors

peci

esian

dzith

ech

arge

num

bero

fspe

ciesiw

here

we

haveK

diffe

rent

spec

ies.

Inth

elim

itof

zero

ioni

cst

reng

th(

κ=

0)eq

ns.4

.18

and

4.19

redu

ceto

the

sim

ple

form

sof

eqns

.4.1

4an

d4.

15re

spec

tivel

y.

4.1.

5M

odifi

edno

n-bo

nded

inte

ract

ions

Inth

eG

RO

MA

CS

forc

efie

ldth

eno

n-bo

nded

pote

ntia

lsca

nbe

mod

ified

bya

shift

func

tion.

The

purp

ose

ofth

isis

tore

plac

eth

etr

unca

ted

forc

esby

forc

esth

atar

eco

ntin

uous

and

have

cont

inuo

usde

rivat

ives

atth

ecu

t-of

frad

ius.

With

such

forc

esth

etim

e-st

epin

tegr

atio

npr

oduc

esm

uch

smal

ler

erro

rsan

dth

ere

are

nosu

chco

mpl

icat

ions

ascr

eatin

gch

arge

sfr

omdi

pole

sby

the

trun

catio

npr

oced

ure.

Infa

ct,b

yus

ing

shift

edfo

rces

ther

eis

none

edfo

rch

arge

grou

psin

the

cons

truc

tion

ofne

ighb

orlis

ts.

How

ever

,th

esh

iftfu

nctio

npr

oduc

esa

cons

ider

able

mod

ifica

tion

ofth

eC

oulo

mb

pote

ntia

l.U

nles

sth

e’m

issi

ng’l

ong-

rang

epo

tent

iali

spr

oper

lyca

lcul

ated

and

adde

d(t

hrou

ghth

eus

eof

PP

PM

,Ew

ald,

orP

ME

),th

eef

fect

ofsu

chm

odifi

catio

nsm

ustb

eca

refu

llyev

alua

ted.

The

mod

ifica

tion

ofth

eLe

nnar

d-Jo

nes

disp

ersi

onan

dre

puls

ion

ison

lym

inor

,but

itdo

esre

mov

eth

eno

ise

caus

edby

cut-

offe

ffect

s.

The

reis

no

fund

amen

tald

iffer

ence

betw

een

asw

itch

func

tion

(whi

chm

ultip

lies

the

pote

ntia

lwith

afu

nctio

n)an

da

shift

func

tion

(whi

chad

dsa

func

tion

toth

efo

rce

orpo

tent

ial).

The

switc

h

Page 72: CS USER AL Simulations - GROMACS

52

Ch

ap

ter

4.

Force

field

s

functionis

aspecialcase

oftheshiftfunction,w

hichw

eapply

tothe

force

fun

ctionF

(r),relatedto

theelectrostatic

orVan

derW

aalsforce

actingon

particlei

byparticle

jas

Fi =

cF(r

ij ) rij

rij

(4.20)

For

pureC

oulomb

orLennard-Jones

interactionsF

(r)=F

α (r)=r −

(α+

1).

The

shiftedforce

Fs (r)

cangenerally

bew

rittenas:

Fs (r)

=F

α (r)r<r1

Fs (r)

=F

α (r)+S

(r)r1≤r<rc

Fs (r)

=0

rc ≤

r

(4.21)

When

r1

=0

thisis

atraditional

shiftfunction,

otherwise

itacts

asa

switch

function.T

hecorresponding

shiftedcoulom

bpotentialthen

reads:

Vs (r

ij )=fΦ

s (rij )q

i qj

(4.22)

whereΦ

(r)is

thepotentialfunctionΦ

s (r)= ∫

∞rF

s (x)dx

(4.23)

The

GR

OM

AC

Sshiftfunction

shouldbe

smooth

attheboundaries,therefore

thefollow

ingbound-

aryconditions

areim

posedon

theshiftfunction:

S(r

1 )=

0S′(r

1 )=

0S

(rc )

=−F

α (rc )

S′(r

c )=

−F′α (r

c )

(4.24)

A3

rd

degreepolynom

ialoftheform

S(r)

=A

(r−r1 )

2+B

(r−r1 )

3(4.25)

fulfillsthese

requirements.

The

constantsA

andB

aregiven

bythe

boundarycondition

atrc :

A=

−(α

+4)r

c−

(α+

1)r1

rα+

2c

(rc −

r1 )

2

B=

(α+

3)rc−

(α+

1)r1

rα+

2c

(rc −

r1 )

3

(4.26)

Thus

thetotalforce

functionis

Fs (r)

=1

rα+

1+A

(r−r1 )

2+B

(r−r1 )

3(4.27)

andthe

potentialfunctionreads

Φ(r)

=1rα−A3

(r−r1 )

3−B4

(r−r1 )

4−C

(4.28)

Appendix

E

ManualP

ages

E.1

options

AllG

RO

MA

CS

programs

have6

standardoptions,ofw

hichsom

eare

hiddenby

default:

Other

options-h

booln

oP

rinthelpinfo

andquit

-Xbool

no

Use

dialogbox

GU

Itoeditcom

mand

lineoptions

-nice

int0

Setthe

nicelevel

•Ifthe

configurationscriptfound

Motifor

Lesstifonyour

system,you

canuse

thegraphicalinterface

(ifnot,youw

illgetanerror):

-Xbooln

oU

sedialog

boxG

UIto

editcomm

andline

options

•W

hencom

piledon

anS

GI-IR

IXsystem

,allGR

OM

AC

Sprogram

shave

anadditionaloption:

-np

riint

0S

etnonblocking

priority(try

128)

•O

ptionalfilesare

notusedunless

theoption

isset,in

contrasttonon

optionalfiles,where

thedefault

filenam

eis

usedw

henthe

optionis

notset.

•A

llGR

OM

AC

Sprogram

sw

illacceptfileoptions

withouta

fileextension

orfilename

beingspecified.

Insuch

casesthe

defaultfilenam

esw

illbe

used.W

ithm

ultipleinput

filetypes,

suchas

genericstructure

format,

thedirectory

will

besearched

forfiles

ofeach

typew

iththe

suppliedor

defaultnam

e.W

henno

suchfile

isfound,or

with

outputfilesthe

firstfiletype

willbe

used.

•A

llGR

OM

AC

Sprogram

sw

iththe

exceptionofmd

run

,n

mru

nand

en

eco

nv

checkif

thecom

-m

andline

optionsare

valid.Ifthis

isnotthe

case,theprogram

willbe

halted.

•E

numerated

options(enum

)shouldbe

usedw

ithone

oftheargum

entslisted

inthe

optiondescription,

theargum

entmay

beabbreviated.

The

firstmatch

tothe

shortestargumentin

thelistw

illbeselected.

•Vector

optionscan

beused

with

1or

3param

eters.W

henonly

oneparam

eteris

suppliedthe

two

othersare

alsosetto

thisvalue.

•F

orm

anyG

RO

MA

CS

programs,the

time

optionscan

besupplied

indifferenttim

eunits,depending

onthe

settingofthe-tu

option.

•A

llGR

OM

AC

Sprogram

scan

readcom

pressedor

g-zippedfiles.

There

might

bea

problemw

ithreading

compressed.xtc

,.trrand

.trjfiles,butthese

willnotcom

pressvery

wellanyw

ay.

Page 73: CS USER AL Simulations - GROMACS

17

6A

pp

en

dix

D.

Ave

rage

sa

nd

fluct

ua

tion

s4

.1.

No

n-b

on

de

din

tera

ctio

ns

53

0.0

1.0

2.0

3.0

4.0

5.0

r�

−0.

5

0.0

0.5

1.0

1.5

f(r)

Nor

mal

For

ceS

hifte

d F

orce

Shi

ft F

unct

ion

Fig

ure

4.4:

The

Cou

lom

bF

orce

,Shi

fted

For

cean

dS

hift

Fun

ctio

nS

(r),

usin

gr 1

=2

and

r c=

4.

whe

re

C=

1 rα c

−A 3

(rc−r 1

)3−B 4

(rc−r 1

)4(4

.29)

Whe

nr 1

=0,

the

mod

ified

Cou

lom

bfo

rce

func

tion

is

Fs(r

)=

1 r2−

5r2

r4 c

+4r

3

r5 c

(4.3

0)

iden

tical

toth

epa

rab

olic

forc

efu

nctio

nre

com

men

ded

tobe

used

asa

shor

t-ra

nge

func

tion

inco

njun

ctio

nw

itha

Poi

sson

solv

erfo

rth

elo

ng-r

ange

part

[13

].T

hem

odifi

edC

oulo

mb

pote

ntia

lfu

nctio

nis

Φ(r

)=

1 r−

5 3rc

+5r

3

3r4 c

−r4 r5 c

(4.3

1)

see

also

Fig

.4.4.

4.1.

6M

odifi

edsh

ort-

rang

ein

tera

ctio

nsw

ithE

wal

dsu

mm

atio

n

Whe

nE

wal

dsu

mm

atio

nor

part

icle

-mes

hE

wal

dis

used

toca

lcul

ate

the

long

-ran

gein

tera

ctio

ns,

the

shor

t-ra

nge

coul

omb

pote

ntia

lmus

tals

obe

mod

ified

,sim

ilar

toth

esw

itch

func

tion

abov

e.In

this

case

the

shor

tran

gepo

tent

iali

sgi

ven

by

V(r

)=f

erfc(βr i

j)

r ij

q iq j,

(4.3

2)

whe

reβ

isa

para

met

erth

atde

term

ines

the

rela

tive

wei

ght

betw

een

the

dire

ctsp

ace

sum

and

the

reci

proc

alsp

ace

sum

and

erfc(x)

isth

eco

mpl

emen

tary

erro

rfu

nctio

n.F

orfu

rthe

rde

tails

onlo

ng-

rang

eel

ectr

osta

tics,

see

sec.

4.6.

Page 74: CS USER AL Simulations - GROMACS

54

Ch

ap

ter

4.

Force

field

s

b0

0.080.09

0.100.11

r (nm)

0 50

100

150

200

V (kJ mole–1

)F

igure4.5:

Principle

ofbondstretching

(left),andthe

bondstretching

potential(right).

4.2B

ondedinteractions

Bonded

interactionsare

basedon

afixed

listof

atoms.

They

arenot

exclusivelypair

interac-tions,

butinclude

3-and

4-bodyinteractions

asw

ell.T

hereare

bo

nd

stretch

ing(2-body),b

on

da

ng

le(3-body),

anddih

ed

rala

ng

le(4-body)interactions.

Aspecialtype

ofdihedralinteraction

(calledim

pro

pe

rd

ihe

dra

l)isused

toforce

atoms

torem

ainin

aplane

orto

preventtransitionto

aconfiguration

ofoppositechirality

(am

irrorim

age).

4.2.1B

ondstretching

Harm

onicpotential

The

bondstretching

between

two

covalentlybonded

atoms

iand

jis

representedby

aharm

onicpotential

Vb(r

ij )=

12k

bij (rij −

bij )

2(4.33)

seealso

Fig.4.5,w

iththe

force

Fi (r

ij )=k

bij (rij −

bij ) r

ij

rij

(4.34)

Fourth

power

potential

Inthe

GR

OM

OS

-96force

field[46]

thecovalent

bondpotentialis

written

forreasons

ofcom

pu-tationalefficiency

as:

Vb(r

ij )=

14k

bij (r2ij −

b2ij )

2(4.35)

thecorresponding

forceis:

Fi (r

ij )=k

bij (r2ij −

b2ij )

rij

(4.36)

D.2

.Im

ple

me

nta

tion

17

5

which

we

canexpand

to:

n∑i=

m S∑s=

1 (xsi )

2+ (

XSm

,n

m−n

+1 )

2−2

XSm

,n

m−n

+1

S∑s=

1

xsi+

S∑s=

1

S∑s ′=

s+1

xsi x

s ′i −

S∑s=

1

n∑i=

m [(xsi )

2−2

Xsm

,n

m−n

+1x

si+ (

Xsm

,n

m−n

+1 )

2 ]=

∆σ

(D.31)

theterm

sw

ith(xsi )

2cancel,so

thatwe

cansim

plifyto:

(X

Sm,n )

2

m−n

+1−

2X

Sm,n

m−n

+1

n∑i=

m

S∑s=

1

xsi −

2n∑i=

m

S∑s=

1

S∑s ′=

s+1

xsi x

s ′i−

S∑s=

1

n∑i=

m [−2

Xsm

,n

m−n

+1x

si+ (

Xsm

,n

m−n

+1 )

2 ]=

∆σ

(D.32)

or

− (X

Sm,n )

2

m−n

+1−

2n∑i=

m

S∑s=

1

S∑s ′=

s+1

xsi x

s ′i+

S∑s=

1 (X

sm,n )

2

m−n

+1

=∆σ

(D.33)

Ifwe

nowexpand

thefirstterm

usingeqn.

D.28

we

obtain:

− (∑Ss=

1X

sm,n )

2

m−n

+1

−2

n∑i=

m

S∑s=

1

S∑s ′=

s+1

xsi x

s ′i+

S∑s=

1 (X

sm,n )

2

m−n

+1

=∆σ

(D.34)

which

we

canreform

ulateto:

−2

S∑s=

1

S∑s ′=

s+1

Xsm

,nX

s ′m

,n+

n∑i=

m

S∑s=

1

S∑s ′=

s+1

xsi x

s ′i =

∆σ

(D.35)

or

−2

S∑s=

1

Xsm

,n

S∑s ′=

s+1

Xs ′m

,n+

S∑s=

1

n∑i=

m

xsi

S∑s ′=

s+1

xs ′i

=∆σ

(D.36)

which

gives

−2

S∑s=

1 X

sm,n

S∑s ′=

s+1

n∑i=

m

xs ′i

+n∑i=

m

xsi

S∑s ′=

s+1

xs ′i

=∆σ

(D.37)

Since

we

needalldata

pointsitoevaluate

this,ingeneralthis

isnotpossible.

We

canthen

make

anestim

ateofσ

Sm,n

usingonly

thedata

pointsthatare

availableusing

thelefthand

sideofeqn.D

.30.W

hilethe

averagecan

becom

putedusing

alltim

esteps

inthe

simulation,

theaccuracy

ofthe

fluctuationsis

thuslim

itedby

thefrequency

with

which

energiesare

saved.S

incethis

canbe

easilydone

with

aprogram

suchas

xmgr

thisis

notbuilt-inin

GR

OM

AC

S.

Page 75: CS USER AL Simulations - GROMACS

17

4A

pp

en

dix

D.

Ave

rage

sa

nd

fluct

ua

tion

s

and

thus n+

m ∑ i=1

[ xA

Bi

−X

AB

1,n

+m

n+m

] 2 =n ∑ i=1

[ xA i−X

A 1,n

n

] 2 +m ∑ i=

1

[ xB i−X

B 1,m

m

] 2 +∆σ

(D.2

4)

or

n+

m ∑ i=1

(xAB

i)2−

2xA

Bi

XA

B1,n

+m

n+m

+

( XA

B1,n

+m

n+m

) 2 −n ∑ i=1

(xA i)2−

2xA i

XA 1,n

n+

( XA 1,n

n

) 2 −m ∑ i=

1

(xB i)2−

2xB i

XB 1,m

m+

( XB 1,m

m

) 2 =∆σ

(D.2

5)

allt

hex

2 ite

rms

drop

out,

and

the

term

sin

depe

nden

toft

hesu

mm

atio

nco

unte

ri

can

besi

mpl

ified

:( X

AB

1,n

+m

) 2n

+m

( XA 1,n

) 2n

( XB 1,m

) 2m

2X

AB

1,n

+m

n+m

n+

m ∑ i=1

xA

Bi

+2X

A 1,n

n

n ∑ i=1

xA i

+2X

B 1,m

m

m ∑ i=1

xB i

=∆σ

(D.2

6)

we

reco

gniz

eth

eth

ree

part

ials

ums

onth

ese

cond

line

and

use

eqn.

D.2

1to

obta

in:

∆σ

=

( mX

A 1,n−nX

B 1,m

) 2nm

(n+m

)(D

.27)

ifw

ech

eck

this

byin

sert

ingm

=1

we

getb

ack

eqn.D

.11

D.2

.3S

umm

ing

ener

gyte

rms

The

gen

ergy

prog

ram

can

also

sum

ener

gyte

rms

into

one,

e.g.

pote

ntia

l+ki

netic

=to

tal.

For

the

part

iala

vera

ges

this

isag

ain

easy

ifw

eha

veS

ener

gyco

mpo

nent

ss:

XS m

,n=

n ∑ i=m

S ∑ s=1

xs i

=S ∑ s=

1

n ∑ i=m

xs i

=S ∑ s=

1

Xs m

,n(D

.28)

For

the

fluct

uatio

nsit

isle

sstr

ivia

laga

in,c

onsi

derin

gfo

rex

ampl

eth

atth

eflu

ctua

tion

inpo

tent

ial

and

kine

ticen

ergy

shou

ldca

ncel

.N

ever

thel

ess

we

can

try

the

sam

eap

proa

chas

befo

reby

writ

ing:

σS m

,n=

S ∑ s=1

σs m

,n+

∆σ

(D.2

9)

ifw

efil

lin

eqn.

D.6

:

n ∑ i=m

[( S ∑ s=1

xs i

) −X

S m,n

m−n

+1

] 2 =S ∑ s=

1

n ∑ i=m

[ (xs i)−

Xs m

,n

m−n

+1

] 2 +∆σ

(D.3

0)

4.2

.B

on

de

din

tera

ctio

ns

55

The

forc

eco

nsta

nts

for

this

form

ofth

epo

tent

iali

sre

late

dto

the

usua

lhar

mon

icfo

rce

cons

tant

kb,

harm

(sec

.4.2

.1)

as2k

bb2 ij

=k

b,harm

(4.3

7)

The

forc

eco

nsta

nts

are

mos

tlyde

rived

from

the

harm

onic

ones

used

inG

RO

MO

S-8

7[

39].

Al-

thou

ghth

isfo

rmis

com

puta

tiona

llym

ore

effic

ient

(bec

ause

nosq

uare

root

has

tobe

eval

uate

d),i

tis

conc

eptu

ally

mor

eco

mpl

ex.

One

part

icul

ardi

sadv

anta

geis

that

sinc

eth

efo

rmis

noth

arm

onic

,th

eav

erag

een

ergy

ofa

sing

lebo

ndis

note

qual

to1 2kT

asit

isfo

rth

eno

rmal

harm

onic

pote

ntia

l.

4.2.

2M

orse

pote

ntia

lbon

dst

retc

hing

For

som

esy

stem

sth

atre

quire

anan

harm

onic

bond

stre

tchi

ngpo

tent

ial,

the

Mor

sepo

tent

ial[

47]

betw

een

two

atom

sian

dj

isav

aila

ble

inG

RO

MA

CS

.T

his

pote

ntia

ldiff

ers

from

the

harm

onic

pote

ntia

lin

havi

ngan

asym

met

ricpo

tent

ialw

ella

nda

zero

forc

eat

infin

itedi

stan

ceT

hefu

nctio

nal

form

is:

Vm

orse

(rij

)=D

ij[1−

exp(−β

ij(r

ij−b i

j))

]2,

(4.3

8)

see

also

Fig

.4.6,

and

the

corr

espo

ndin

gfo

rce

is:

Fm

orse

(rij)

=2D

ijβ

ijr i

jex

p(−β

ij(r

ij−b i

j))∗

[1−

exp(−β

ij(r

ij−b i

j))

]r ij

r ij,

(4.3

9)

whe

reD

ijis

the

dept

hof

the

wel

lin

kJ/m

ol,βij

defin

esth

est

eepn

ess

ofth

ew

ell(

innm−1),

and

b ij

isth

eeq

uilib

rium

dist

ance

innm

.T

hest

eepn

ess

para

met

erβ

ijca

nbe

expr

esse

din

term

sof

the

redu

ced

mas

sof

the

atom

sia

ndj,

the

fund

amen

talv

ibra

tion

freq

uenc

ijan

dth

ew

elld

epth

Dij

:

βij

ij

√ µij

2Dij

(4.4

0)

and

beca

useω

=√ k

/µ,o

neca

nre

writ

eβij

inte

rms

ofth

eha

rmon

icfo

rce

cons

tantk ij

βij

=

√ kij

2Dij

(4.4

1)

For

smal

ldev

iatio

ns(r

ij−b i

j),

one

can

expa

ndth

eexp-

term

tofir

st-o

rder

inth

eTa

ylor

expa

nsio

n:

exp(−x)≈

1−x

(4.4

2)

Sub

stitu

ting

this

inth

efu

nctio

nalf

rom

;

Vm

orse

(rij)

=D

ij[1−

exp(−β

ij(r

ij−b i

j))

]2

=D

ij[1−

(1−√ k

ij

2D

ij(r

ij−b i

j))

]2

=1 2k

ij(r

ij−b i

j))

2,

(4.4

3)

one

reco

vers

the

harm

onic

bond

stre

tchi

ngpo

tent

ial.

Page 76: CS USER AL Simulations - GROMACS

56

Ch

ap

ter

4.

Force

field

s

0.10.2

0.30.4

0.50.6

rij (nm)

0

100

200

300

400

Vij (kJ / mol)

Figure

4.6:T

heM

orsepotentialw

ell,with

bondlength

0.15nm

.

4.2.3C

ubicbond

stretchingpotential

Another

anharmonic

bondstretching

potentialthat

isslightly

simpler

thanthe

Morse

potentialadds

acubic

termin

thedistance

tothe

simple

harmonic

form:

Vb(r

ij )=k

bij (rij −

bij )

2+k

bij kcu

bij

(rij −

bij )

3(4.44)

Aflexible

water

model(based

onthe

SP

Cw

aterm

odel[48])

includinga

cubicbond

stretchingpotential

forthe

O-H

bondw

asdeveloped

byF

erguson[

49].T

hism

odelw

asfound

toyield

areasonable

Infraredspectrum

.T

heF

ergusonw

aterm

odelisavailable

inthe

GR

OM

AC

Slibrary.

Itshould

benoted

thatthe

potentialisasym

etric,overstretching

leadsto

infinitelylow

energies.T

heintegration

timestep

istherefore

limited

to1

fs.

The

forcecorresponding

tothis

potentialis:

Fi (r

ij )=

2k

bij (rij −

bij )

rij

rij

+3k

bij kcu

bij

(rij −

bij )

2r

ij

rij

(4.45)

4.2.4H

armonic

anglepotential

The

bondangle

vibrationbetw

eena

tripletof

atoms

i-j

-k

isalso

representedby

aharm

onicpotentialon

theangleθ

ijk

Va (θ

ijk )=

12k

θijk (θijk

−θ0ijk )

2(4.46)

As

thebond-angle

vibrationis

representedby

aharm

onicpotentialthe

formis

thesam

eas

thebond

stretching(F

ig.4.5).

The

forceequations

aregiven

bythe

chainrule:

Fi

=−dV

a (θijk )

dr

i

Fk

=−dV

a (θijk )

dr

kF

j=

−F

i −F

k

where

θijk

=arccos

(rij ·r

kj )

rij r

kj

(4.47)

D.2

.Im

ple

me

nta

tion

17

3

D.2.1

Partofa

Sim

ulation

Itis

notuncom

mon

toperform

asim

ulationw

herethe

firstpart,

e.g.100

ps,is

takenas

equili-bration.

How

ever,the

averagesand

fluctuationsas

printedin

thelog

fileare

computed

overthe

whole

simulation.

The

equilibrationtim

e,which

isnow

partofthesim

ulation,may

insuch

acase

invalidatethe

averagesand

fluctuations,because

thesenum

bersare

nowdom

inatedby

theinitial

drifttowards

equilibrium.

Using

eqns.D.7

andD

.8the

averageand

standarddeviation

overpart

ofthe

trajectorycan

becom

putedas:X

m+

1,m

+k

=X

1,m

+k−X

1,m

(D.15)

σm

+1,m

+k

1,m

+k−σ

1,m− [

X1,m

m−X

1,m

+k

m+k ]

2m

(m+k)

k(D

.16)

or,more

generally(w

ithp≥

1and

q≥p):

Xp,q

=X

1,q −

X1,p−

1(D

.17)

σp,q

1,q −

σ1,p−

1 − [X

1,p−

1

p−

1−X

1,q

q ]2

(p−

1)qq−p

+1

(D.18)

Note

thatim

plementation

ofthis

isnot

entirelytrivial,

sinceenergies

arenot

storedevery

time

stepofthe

simulation.

We

thereforehave

toconstruct

X1,p−

1and

σ1,p−

1from

theinform

ationat

timep

usingeqns.D

.11and

D.12:

X1,p−

1=

X1,p −

xp

(D.19)

σ1,p−

1=

σ1,p −

[X

1,p−

1 −(p−

1)xp

] 2

(p−

1)p(D

.20)

D.2.2

Com

biningtw

osim

ulations

Another

frequentlyoccurring

problemis,

thatthe

fluctuationsof

two

simulations

must

becom

-bined.

Consider

thefollow

ingexam

ple:w

ehave

two

simulations

(A)

ofn

and(B

)ofm

steps,inw

hichthe

secondsim

ulationis

acontinuation

ofthe

first.H

owever,

thesecond

simulation

startsnum

beringfrom

1instead

offromn

+1.

For

thepartialsum

thisis

noproblem

,w

ehave

toadd

XA1,n

fromrun

A:

XA

B1,n

+m

=X

A1,n

+X

B1,m

(D.21)

When

we

want

tocom

putethe

partialvariance

fromthe

two

components

we

haveto

make

acorrection∆

σ:

σA

B1,n

+m

A1,n

B1,m

+∆σ

(D.22)

ifwe

definex

AB

ias

thecom

binedand

renumbered

setofdatapoints

we

canw

rite:

σA

B1,n

+m

=n+

m∑i=

1 [x

AB

i−X

AB

1,n

+m

n+m ]

2

(D.23)

Page 77: CS USER AL Simulations - GROMACS

17

2A

pp

en

dix

D.

Ave

rage

sa

nd

fluct

ua

tion

s

and

the

part

ialv

aria

nce

σn,m

=m ∑ i=

n

[ xi−

Xn,m

m−n

+1

] 2(D

.6)

Itca

nbe

show

nth

atX

n,m

+k

=X

n,m

+X

m+

1,m

+k

(D.7

)

and

σn,m

+k

n,m

m+

1,m

+k

+[

Xn,m

m−n

+1−

Xn,m

+k

m+k−n

+1

] 2 ∗

(m−n

+1)

(m+k−n

+1)

k(D

.8)

Forn

=1

one

finds

σ1,m

+k

1,m

m+

1,m

+k

+[ X 1

,m

m−X

1,m

+k

m+k

] 2 m(m

+k)

k(D

.9)

and

forn

=1

andk

=1

(eqn

.D.8

)be

com

es

σ1,m

+1

1,m

+[ X 1

,m

m−X

1,m

+1

m+

1

] 2 m(m

+1)

(D.1

0)

1,m

+[X

1,m−mx

m+

1]2

m(m

+1)

(D.1

1)

whe

rew

eha

veus

edth

ere

latio

n X1,m

+1

=X

1,m

+x

m+

1(D

.12)

Usi

ngfo

rmul

ae(e

qn.D.1

1)an

d(e

qn.D

.12)

the

aver

age

〈x〉

=X

1,N

x

Nx

(D.1

3)

and

the

fluct

uatio

n⟨ (∆

x)2⟩1 2

=[ σ 1,

Nx

Nx

]1 2

(D.1

4)

can

beob

tain

edby

one

swee

pth

roug

hth

eda

ta.

D.2

Impl

emen

tatio

n

InG

RO

MA

CS

the

inst

anta

neou

sen

ergi

esE(m

)are

stor

edin

the

ener

gyfil

e,al

ong

with

the

valu

esofσ

1,m

andX

1,m

.A

lthou

ghth

est

eps

are

coun

ted

from

0,fo

rthe

ener

gyan

dflu

ctua

tions

step

sar

eco

unte

dfr

om1.

Thi

sm

eans

that

the

equa

tions

pres

ente

dhe

rear

eth

eon

esth

atar

eim

plem

ente

d.W

egi

veso

mew

hatl

engt

hyde

rivat

ions

inth

isse

ctio

nto

sim

plify

chec

king

ofco

dean

deq

uatio

nsla

ter

on.

4.2

.B

on

de

din

tera

ctio

ns

57

θ 0

100.

0

110.

0

120.

0

130.

0

140.

0

θ

0.0

10.0

20.0

30.0

40.0

50.0

Va (kJ mole �

–1)

Fig

ure

4.7:

Prin

cipl

eof

angl

evi

brat

ion

(left)

and

the

bond

angl

epo

tent

ial(

right

).

The

num

berin

gi,j,k

isin

sequ

ence

ofco

vale

ntly

bond

edat

oms,

withjde

notin

gth

em

iddl

eat

om(s

eeF

ig.4

.7).

4.2.

5C

osin

eba

sed

angl

epo

tent

ial

Inth

eG

RO

MO

S-9

6fo

rce

field

asi

mpl

ified

func

tion

isus

edto

repr

esen

tang

levi

brat

ions

:

Va(θ

ijk)

=1 2k

θ ijk

( cos(θ i

jk)−

cos(θ0 ij

k)) 2

(4.4

8)

whe

reco

s(θ i

jk)

=r

ij·r

kj

r ijr k

j(4

.49)

The

corr

espo

ndin

gfo

rce

can

bede

rived

bypa

rtia

ldiff

eren

tiatio

nw

ithre

spec

tto

the

atom

icpo

si-

tions

.T

hefo

rce

cons

tant

sin

this

func

tion

are

rela

ted

toth

efo

rce

cons

tant

sin

the

harm

onic

form

kθ,h

arm

(sec

.4.2

.4)

by:

kθsi

n2(θ

0 ijk)

=k

θ,h

arm

(4.5

0)

4.2.

6Im

prop

erdi

hedr

als

Impr

oper

dihe

dral

sar

em

eant

toke

eppl

anar

grou

pspl

anar

(e.g

.ar

omat

icrin

gs)

orto

prev

ent

mol

ecul

esfr

omfli

ppin

gov

erto

thei

rm

irror

imag

es,s

eeF

ig.

4.8.

Vid

(ξij

kl)

=k

ξ(ξ

ijkl−ξ 0

)2(4

.51)

Thi

sis

also

aha

rmon

icpo

tent

ial,

itis

plot

ted

inF

ig.

4.9.

Not

eth

at,s

ince

itis

harm

onic

,per

iodi

city

isno

tta

ken

into

acco

unt,

soit

isbe

stto

defin

eim

prop

erdi

hedr

als

toha

vea

ξ 0as

far

away

from

±18

0◦as

you

can

man

age.

4.2.

7P

rope

rdi

hedr

als

For

the

norm

aldi

hedr

alin

tera

ctio

nth

ere

isa

choi

ceof

eith

erth

eG

RO

MO

Spe

riodi

cfu

nctio

nor

afu

nctio

nba

sed

onex

pans

ion

inpo

wer

sofcosφ

(the

so-c

alle

dR

ycka

ert-

Bel

lem

ans

pote

ntia

l).T

his

Page 78: CS USER AL Simulations - GROMACS

58

Ch

ap

ter

4.

Force

field

s

k

li

j

i

kj

l

k

i

j

l

Figure

4.8:P

rincipleof

improper

dihedralangles.

Out

ofplane

bendingfor

rings(left),

sub-stituents

ofrings

(middle),

outoftetrahedral(right).T

heim

properdihedralangle

ξis

definedas

theangle

between

planes(i,j,k)

and(j,k,l)

inallcases.

–20.0–10.0

0.010.0 �

20.0 �

ξ

0.0

10.0

20.0

30.0

Vi (kJ mole�–1

)

Figure

4.9:Im

properdihedralpotential.

Appendix

D

Averages

andfluctuations

D.1

Form

ulaefor

averaging

Note:this

sectionw

astaken

fromref[

77].

When

analyzinga

MD

trajectoryaverages

〈x〉and

fluctuations

⟨(∆x)

2 ⟩12

= ⟨[x−〈x〉] 2 ⟩

12(D

.1)

ofaquantityx

areto

becom

puted.T

hevariance

σx

ofaseries

ofNxvalues,{x

i },canbe

computed

from

σx

=N

x∑i=

1

x2i−

1Nx (

Nx

∑i=1

xi )

2

(D.2)

Unfortunately

thisform

ulais

numerically

notveryaccurate,especially

whenσ

12xis

smallcom

paredto

thevalues

ofxi .

The

following

(equivalent)expression

isnum

ericallym

oreaccurate

σx

=N

x∑i=

1 [xi −

〈x〉] 2(D

.3)

with

〈x〉=

1Nx

Nx

∑i=1

xi

(D.4)

Using

eqns.D.2

andD

.4one

hasto

gothrough

theseries

ofx

ivalues

twice,

onceto

determine

〈x〉and

againto

computeσ

x ,w

hereaseqn.D.1

requiresonly

onesequential

scanof

theseries

{xi }.

How

ever,onem

aycasteqn.

D.2

inanother

form,containing

partialsums,w

hichallow

sfor

asequentialupdate

algorithm.

Define

thepartialsum

Xn,m

=m∑i=

n

xi

(D.5)

Page 79: CS USER AL Simulations - GROMACS

17

0A

pp

en

dix

C.

Lo

ng

ran

geco

rre

ctio

ns

4.2

.B

on

de

din

tera

ctio

ns

59

j

k

l

i

0.0

90.0�

180.

0

270.

0

360.

0

φ

0.0

20.0

40.0

60.0

80.0

Vd (kJ mole �

–1)

Fig

ure

4.10

:P

rinci

ple

ofpr

oper

dihe

dral

angl

e(le

ft,intra

nsf

orm

)and

the

dihe

dral

angl

epo

tent

ial

(rig

ht).

C0

9.28

C2

-13.

12C

426

.24

C1

12.1

6C

3-3

.06

C5

-31.

5

Tabl

e4.

1:C

onst

ants

for

Ryc

kaer

t-B

elle

man

spo

tent

ial(

kJm

ol−

1).

choi

ceha

sco

nseq

uenc

esfo

rth

ein

clus

ion

ofsp

ecia

lint

erac

tions

betw

een

the

first

and

the

four

that

omof

the

dihe

dral

quad

rupl

e.W

ithth

epe

riodi

cG

RO

MO

Spo

tent

iala

spec

ial1

-4LJ

-inte

ract

ion

mus

tbe

incl

uded

;w

ithth

eR

ycka

ert-

Bel

lem

ans

pote

ntia

lthe

1-4

inte

ract

ions

mus

tbe

excl

uded

from

the

non-

bond

edlis

t.

Pro

per

dihe

dral

s:pe

riodi

cty

pe

Pro

per

dihe

dral

angl

esar

ede

fined

acco

rdin

gto

the

IUP

AC

/IUB

conv

entio

n,w

here

φis

the

angl

ebe

twee

nth

eijk

and

thejkl

plan

es,

with

zero

corr

espo

ndin

gto

theci

sco

nfigu

ratio

n(i

andl

onth

esa

me

side

).

Vd(φ

ijkl)

=k

φ(1

+co

s(nφ−φ

0))

(4.5

2)

Pro

per

dihe

dral

s:R

ycka

ert-

Bel

lem

ans

func

tion

For

alka

nes,

the

follo

win

gpr

oper

dihe

dral

pote

ntia

lis

ofte

nus

ed(s

eeF

ig.

4.11

)

Vrb(φ

ijkl)

=5 ∑ n=

0

Cn(c

os(ψ

))n,

(4.5

3)

whe

reψ

=φ−

180◦

.N

ote:

Aco

nver

sion

from

one

conv

entio

nto

anot

her

can

beac

hiev

edby

mul

tiply

ing

ever

yco

effi-

cien

tCn

by(−

1)n.

An

exam

ple

ofco

nsta

nts

forC

isgi

ven

inTa

ble4

.1.

Page 80: CS USER AL Simulations - GROMACS

60

Ch

ap

ter

4.

Force

field

s

0.090.0 �

180.0 �

270.0 �

360.0 �

φ

0.0

10.0

20.0

30.0

40.0

50.0

Vd (kJ mole�–1

)

Figure

4.11:R

yckaert-Bellem

ansdihedralpotential.

(Note:T

heuse

ofthispotentialim

pliesexclusions

ofLJ-interactionsbetw

eenthe

firstandthe

lastatom

ofthedihedral,andψ

isdefined

accordingto

the’polym

erconvention’(

ψtr

ans

=0).)

The

RB

dihedralfunctioncan

alsobe

usedto

includethe

OP

LSdihedralpotential[

50].T

heO

PLS

potentialfunctionis

givenas

thefirstfour

terms

ofaF

ourierseries:

Vrb (φ

ijkl )

=V

0+

12(V

1 (1+

cos(ψ))

+V

2 (1−

cos(2ψ))

+V

3 (1+

cos(3ψ))),

(4.54)

with

ψ=

φ(protein

convention).B

ecauseof

theequalities

cos(2φ)=

2(cos(φ))2−

1and

cos(3φ)=

4(cos(φ))3−

3cos(φ),one

cantranslate

theO

PLS

parameters

toR

yckaert-Bellem

ansparam

etersas

follows:

C0

=V

0+V

2+

12 (V1+V

3 )C

1=

12 (3V3 −

V1 )

C2

=−V

2

C3

=−

2V

3

C4

=0

C5

=0

(4.55)

with

OP

LSparam

etersin

proteinconvention

andR

Bparam

etersin

polymer

convention.N

ote:Mind

theconversion

fromkcalm

ol −

1forO

PLS

andR

Bparam

etersin

literatureto

kJm

ol −

1

inG

RO

MA

CS

.

4.2.8S

pecialinteractions

Special

potentialsare

usedfor

imposing

restraintson

them

otionof

thesystem

,either

toavoid

disastrousdeviations,or

toinclude

knowledge

fromexperim

entaldata.In

eithercase

theyare

notreally

partof

theforce

fieldand

thereliability

ofthe

parameters

isnot

important.

The

potentialform

s,asim

plemented

inG

RO

MA

CS

,arem

entionedjustfor

thesake

ofcompleteness.

C.1

.D

ispe

rsion

16

9

For

homogeneous

mixtures

we

canagain

usethe

averagedispersion

constant〈C

6 〉(eqn.C

.6):

Plr

=−

43π〈C

6 〉ρ2r −

3c

(C.12)

For

inhomogeneous

systems

eqn.C

.12can

beapplied

underthe

same

restrictionas

holdsfor

theenergy

(seesec.C.1.1).

Page 81: CS USER AL Simulations - GROMACS

16

8A

pp

en

dix

C.

Lo

ng

ran

geco

rre

ctio

ns

Ifw

eco

nsid

erfo

rex

ampl

ea

box

ofpu

rew

ater

,si

mul

ated

with

acu

t-of

fof0

.9nm

and

ade

nsity

of1

gcm

−3

this

corr

ectio

nis

-0.2

5kJ

mol−1.

For

aho

mog

eneo

usm

ixtu

reofM

com

pone

ntsj

with

Nj

part

icle

sea

ch,

we

can

writ

eth

elo

ngra

nge

cont

ribut

ion

toth

een

ergy

as:

Vlr

=M ∑ i6=

j

−2N

iNj

3VπC

6(ij)r−

3c

(C.5

)

Thi

sca

nbe

rew

ritte

nif

we

defin

eana

vera

ged

isp

ers

ion

con

sta

nt〈C6〉:

〈C6〉

=∑ i6=

j

NiN

j

N2C

6(ij)

(C.6

)

Vlr

=−

2 3Nρπ〈C

6〉r

−3

c(C

.7)

Asp

ecia

lfor

mof

ano

n-ho

mog

eneo

ussy

stem

inth

isre

spec

t,is

apu

reliq

uid

inw

hich

the

atom

sha

vedi

ffere

ntC

6va

lues

.In

prac

tice

this

defin

ition

enco

mpa

sses

alm

ost

ever

ym

olec

ule,

exce

ptm

ono-

atom

icm

olec

ules

and

sym

met

ricm

olec

ules

like

N2

orO

2.

The

refo

rew

eal

way

sha

veto

dete

rmin

eth

eav

erag

edi

sper

sion

cons

tant

〈C6〉i

nsi

mul

atio

ns.

Inth

eca

seof

inho

mog

eneo

ussi

mul

atio

nsy

stem

s,e.

g.a

syst

emw

itha

lipid

inte

rfac

e,th

een

ergy

corr

ectio

nca

nbe

appl

ied

if〈C6〉f

orbo

thco

mpo

nent

sis

com

para

ble.

C.1

.2V

irial

and

pres

sure

The

scal

arvi

rialo

fth

esy

stem

due

toth

edi

sper

sion

inte

ract

ion

betw

een

two

part

icle

si

andj

isgi

ven

by:

Ξ=

−r

ij·F

ij=

6C6r−

6ij

(C.8

)

The

pres

sure

isgi

ven

by:

P=

2 3V

(Ekin−

Ξ)

(C.9

)

We

can

agai

nin

tegr

ate

the

long

rang

eco

ntrib

utio

nto

the

viria

l[65

]:

Ξlr

=1 2Nρ

∫ ∞ r c4πr2

Ξdr

=12NπρC

6

∫ ∞ r cr−

4ij

dr

=4πC

6Nρr−

3c

(C.1

0)

The

corr

espo

ndin

gco

rrec

tion

toth

epr

essu

reis

Plr

=−

4 3πC

6ρ2r−

3c

(C.1

1)

Usi

ngth

esa

me

exam

ple

ofa

wat

erbo

x,th

eco

rrec

tion

toth

evi

riali

s3

kJm

ol−

1th

eco

rres

pond

ing

corr

ectio

nto

the

pres

sure

for

SP

Cw

ater

atliq

uid

dens

ityis

appr

ox.

-280

bar.

4.2

.B

on

de

din

tera

ctio

ns

61

0.00�

0.02�

0.04�

0.06�

0.08�

0.10�

r-R

(nm

)�

0.0

2.0

4.0

6.0

8.0

10.0

Vposre (kJ mole–1

)

Fig

ure

4.12

:P

ositi

onre

stra

intp

oten

tial.

4.2.

9P

ositi

onre

stra

ints

The

sear

eus

edto

rest

rain

part

icle

sto

fixed

refe

renc

epo

sitio

nsR

i.T

hey

can

beus

eddu

ring

equi

libra

tion

inor

dert

oav

oid

too

dras

ticre

arra

ngem

ents

ofcr

itica

lpar

ts(e

.g.t

ore

stra

inm

otio

nin

apr

otei

nth

atis

subj

ecte

dto

larg

eso

lven

tfor

ces

whe

nth

eso

lven

tis

noty

eteq

uilib

rate

d).

Ano

ther

appl

icat

ion

isth

ere

stra

inin

gof

part

icle

sin

ash

ella

roun

da

regi

onth

atis

sim

ulat

edin

deta

il,w

hile

the

shel

lis

only

appr

oxim

ated

beca

use

itla

cks

prop

erin

tera

ctio

nfr

omm

issi

ngpa

rtic

les

outs

ide

the

shel

l.R

estr

aini

ngw

illth

enm

aint

ain

the

inte

grity

ofth

ein

ner

part

.F

orsp

heric

alsh

ells

itis

aw

ise

proc

edur

eto

mak

eth

efo

rce

cons

tant

depe

ndon

the

radi

us,i

ncre

asin

gfr

omze

roat

the

inne

rbo

unda

ryto

ala

rge

valu

eat

the

oute

rbo

unda

ry.

Thi

sap

plic

atio

nha

sno

tbe

enim

plem

ente

din

GR

OM

AC

Sho

wev

er.

The

follo

win

gfo

rmis

used

:

Vpr(r

i)=

1 2k

pr|r

i−

Ri|2

(4.5

6)

The

pote

ntia

lis

plot

ted

inF

ig.4.12

.

The

pote

ntia

lfor

mca

nbe

rew

ritte

nw

ithou

tlos

sof

gene

ralit

yas

:

Vpr(r

i)=

1 2

[ kx pr(x

i−X

i)2x

+k

y pr(y

i−Y

i)2y

+k

z pr(z

i−Z

i)2z]

(4.5

7)

Now

the

forc

esar

e:F

x i=

−k

x pr

(xi−X

i)F

y i=

−k

y pr

(yi−Y

i)F

z i=

−k

z pr

(zi−Z

i)(4

.58)

Usi

ngth

ree

diffe

rent

forc

eco

nsta

nts

the

posi

tion

rest

rain

tsca

nbe

turn

edon

orof

fin

each

spat

ial

dim

ensi

on;

this

mea

nsth

atat

oms

can

beha

rmon

ical

lyre

stra

ined

toa

plan

eor

alin

e.P

ositi

onre

stra

ints

are

appl

ied

toa

spec

ial

fixed

list

ofat

oms.

Suc

ha

list

isus

ually

gene

rate

dby

the

pdb2

gmx

prog

ram

.

Page 82: CS USER AL Simulations - GROMACS

62

Ch

ap

ter

4.

Force

field

s

4.2.10A

nglerestraints

These

areused

torestrain

theangle

between

two

pairsofparticles

orbetw

eenone

pairofparticles

andthe

Z-axis.

The

functionalformis

similar

tothatofa

properdihedral.

For

two

pairsofatom

s:

Var (r

i ,rj ,r

k ,rl )

=k

ar (1

−cos(n(θ

−θ0 ))),

where

θ=

arccos (r

j −r

i

‖r

j −r

i ‖·

rl −

rk

‖r

l −r

k ‖ )(4.59)F

orone

pairofatom

sand

theZ

-axis:

Var (r

i ,rj )

=k

ar (1

−cos(n(θ

−θ0 ))),

where

θ=

arccos r

j −r

i

‖r

j −r

i ‖·

001 (4.60)

Am

ultiplicity(n

)of

2is

usefulwhen

youdo

notw

antto

distinguishbetw

eenparalleland

anti-parallelvectors.

4.2.11D

istancerestraints

Distance

restraintsadd

apenalty

tothe

potentialw

henthe

distancebetw

eenspecified

pairsof

atoms

exceedsa

thresholdvalue.

They

arenorm

allyused

toim

poseexperim

entalrestraints,as

fromexperim

entsin

nuclearm

agneticresonance

(NM

R),on

them

otionofthe

system.

Thus

MD

canbe

usedfor

structurerefinem

entusing

NM

Rdata.

The

potentialform

isquadratic

belowa

specifiedlow

erbound

andbetw

eentw

ospecified

upperbounds

andlinear

beyondthe

largestbound

(seeF

ig.4.13).

Vdr (r

ij )=

12 kdr (r

ij −r0 )

2for

rij

<r0

0for

r0

≤rij

<r1

12 kdr (r

ij −r1 )

2for

r1

≤rij

<r2

12 kdr (r

2 −r1 )(2r

ij −r2 −

r1 )

forr2

≤rij

(4.61)

The

forcesare

Fi

=

−k

dr (r

ij −r0 ) r

ij

rij

forrij

<r0

0for

r0

≤rij

<r1

−k

dr (r

ij −r1 ) r

ij

rij

forr1

≤rij

<r2

−k

dr (r

2 −r1 ) r

ij

rij

forr2

≤rij

(4.62)

Tim

eaveraging

Distance

restraintsbased

oninstantaneous

distancescan

potentiallyreduce

thefluctuations

ina

molecule

significantly.T

hisproblem

canbe

overcome

byrestraining

toa

time

ave

rageddis-

Appendix

C

Longrange

corrections

C.1

Dispersion

Inthis

sectionw

ederive

longrange

correctionsdue

tothe

useof

acut-off

forLennard

Jonesinteractions.

We

assume

thatthecut-offis

solong

thattherepulsion

termcan

safelybe

neglected,and

thereforeonly

thedispersion

termis

takeninto

account.D

ueto

thenature

ofthe

dispersioninteraction,

energyand

pressurecorrections

bothare

negative.W

hilethe

energycorrection

isusually

small,itm

aybe

importantforfree

energycalculations.

The

pressurecorrection

incontrast

isvery

largeand

cannotbe

neglected.A

lthoughitis

inprinciple

possibleto

parameterize

aforce

fieldsuch

thatthe

pressureis

closeto

1bar

evenw

ithoutcorrection,

sucha

method

makes

theparam

eterizationdependent

onthe

cut-offand

istherefore

undesirable.P

leasenote

thatit

isnot

consistentto

usethe

longrange

correctionto

thedispersion

without

usingeither

areaction

fieldm

ethodor

aproper

longrange

electrostaticsm

ethodsuch

asE

wald

summ

ationor

PP

PM

.

C.1.1

Energy

The

longrange

contributionofthe

dispersioninteraction

tothe

virialcanbe

derivedanalytically,if

we

assume

ahom

ogeneoussystem

beyondthe

cut-offdistancerc .

The

dispersionenergy

between

two

particlesis

written

as:V

(rij )

=−C

6 r −6

ij(C

.1)

andthe

correspondingforce

isF

ij=

−6C

6 r −8

ijr

ij(C

.2)

The

longrange

contributionto

thedispersion

energyin

asystem

with

Nparticles

andparticle

densityρ=N/V

,whereV

isthe

volume,is

[65]:

Vlr

=12Nρ ∫

∞rc

4πr2g(r)V

(r)dr

(C.3)

which

we

canintegrate

assuming

thattheradialdistribution

functiong(r)

is1

beyondthe

cut-offrc

Vlr

=−

23NρπC

6 r −3

c(C

.4)

Page 83: CS USER AL Simulations - GROMACS

16

6A

pp

en

dix

B.

So

me

imp

lem

en

tatio

nd

eta

ils

elec

tros

tatic

,dis

pers

ion

and

repu

lsio

nin

tera

ctio

ns,b

utfo

rth

esa

keof

cach

ing

perf

orm

ance

thes

eha

vebe

enco

mbi

ned

into

asi

ngle

arra

y.T

hecu

bic

splin

ein

terp

olat

ion

look

slik

eth

is:

y(x

)=

ηy i

+εy

i+1+h

2 6

[ (η3−η)y

′′ i+

(ε3−ε)y′′ i+

1

](B

.58)

whe

reε

=1-η,a

ndy i

andy′′ i

are

the

tabu

late

dva

lues

ofa

func

tiony(x

)an

dits

seco

ndde

rivat

ive

resp

ectiv

ely.

Fur

ther

mor

e,

h=

xi+

1−x

i(B

.59)

ε=

(x−x

i)/h

(B.6

0)

soth

at0≤ε<

1.eq

n.B

.58

can

bere

writ

ten

as

y(x

)=

y i+ε

( y i+

1−y i−h

2 6

( 2y′′ i

+y′′ i+

1

)) +ε2( h2 2

y′′ i

) +ε3h

2 6

( y′′ i+

1−y′′ i

) (B.6

1)

Not

eth

atth

ex-

depe

nden

ceis

com

plet

ely

inε.

Thi

sca

nab

brev

iate

dto

y(x

)=

y i+εF

i+ε2G

i+ε3H

i(B

.62)

Fro

mth

isw

eca

nca

lcul

ate

the

deriv

ativ

ein

orde

rto

dete

rmin

eth

efo

rces

:

dy(x

)dx

=dy

(x)

dεdε dx

=(F

i+

2εG

i+

3ε2H

i)/h

(B.6

3)

Ifw

est

ore

inth

eta

bley

i,F

i,G

ian

dH

iw

ene

eda

tabl

eof

leng

th4n

.T

henu

mbe

rof

poin

tspe

rna

nom

eter

shou

ldbe

onth

eor

der

of50

0to

1000

,fo

rac

cura

tere

pres

enta

tion

(rel

ativ

eer

ror

<10−

4w

hen

n=

500

poin

ts/n

m).

The

forc

ero

utin

esge

tasc

alin

gfa

ctor

sas

apa

ram

eter

that

iseq

ualt

oth

enu

mbe

rof

poin

tspe

rnm

.(N

ote

thath

iss−

1).

The

algo

rithm

goes

alit

tleso

met

hing

like

this

:

1.C

alcu

late

dist

ance

vect

or(

rij

)an

ddi

stan

cer ij

2.M

ultip

lyr i

jbys

and

trun

cate

toan

inte

ger

valu

en

0to

geta

tabl

ein

dex

3.C

alcu

late

frac

tiona

lcom

pone

nt(

ε=sr

ij−n

0)

andε2

4.D

oth

ein

terp

olat

ion

toca

lcul

ate

the

pote

ntia

lV

and

the

the

scal

arfo

rcef

5.C

alcu

late

the

vect

orfo

rceF

bym

ultip

lyin

gf

with

rij

The

tabl

esar

est

ored

inte

rnal

lyas

yi,

F i,

Gi,

Hi

inth

eor

der

coul

omb,

disp

ersi

on,

repu

lsio

n.In

tota

lthe

rear

e12

valu

esin

each

tabl

een

try.

Not

eth

atta

ble

look

upis

sign

ifica

ntly

slo

we

rtha

nco

mpu

tatio

nof

the

mos

tsi

mpl

eLe

nnar

d-Jo

nes

and

Cou

lom

bin

tera

ctio

n.H

owev

er,

itis

muc

hfa

ster

than

the

shift

edco

ulom

bfu

nctio

nus

edin

conj

unct

ion

with

the

PP

PM

met

hod.

Fin

ally

itis

muc

hea

sier

tom

odify

ata

ble

for

the

pote

ntia

l(an

dge

ta

grap

hica

lrep

rese

ntat

ion

ofit)

than

tom

odify

the

inne

rlo

ops

ofth

eM

Dpr

ogra

m.

4.2

.B

on

de

din

tera

ctio

ns

63

0�

0.1

0.2

0.3

0.4

0.5

r (n

m)

051015

Vdisre (kJ mol−1

)

r 0r 1

r 2

Fig

ure

4.13

:D

ista

nce

Res

trai

ntpo

tent

ial.

tanc

e[51

].T

hefo

rces

with

time

aver

agin

gar

e:

Fi

=

−k

dr(r

ij−r 0

)rij

r ij

for

r ij

<r 0

0fo

rr 0

≤r i

j<

r 1

−k

dr(r

ij−r 1

)rij

r ij

forr 1

≤r i

j<

r 2

−k

dr(r

2−r 1

)rij

r ij

forr 2

≤r i

j

(4.6

3)

whe

rer i

jis

give

nby

:

r ij

=<r−

3ij

>−

1/3

(4.6

4)

Bec

ause

ofth

etim

eav

erag

ing

we

can

nolo

nger

spea

kof

adi

stan

cere

stra

intp

oten

tial.

Thi

sw

ayan

atom

can

satis

fytw

oin

com

patib

ledi

stan

cere

stra

ints

on

ave

rage

bym

ovin

gbe

twee

ntw

opo

sitio

ns.

An

exam

ple

wou

ldbe

anam

ino-

acid

side

-cha

inw

hich

isro

tatin

gar

ound

itsχ

dihe

dral

angl

e,th

ereb

yco

min

gcl

ose

tova

rious

othe

rgr

oups

.S

uch

am

obile

side

chai

nm

aygi

veris

eto

mul

tiple

NO

Es,

whi

chca

nno

tbe

fulfi

lled

ina

sing

lest

ruct

ure.

The

com

puta

tion

ofth

etim

eav

erag

eddi

stan

cein

the

md

run

prog

ram

isdo

nein

the

follo

win

gfa

shio

n:

r−3ij(0

)=

r ij(0

)−3

r−3ij(t

)=

r−3ij(t−

∆t)

exp( −

∆t

τ

) +r i

j(t

)−3[ 1−

exp( −

∆t

τ

)](4

.65)

Whe

na

pair

isw

ithin

the

boun

dsit

can

still

feel

afo

rce,

beca

use

the

time

aver

aged

dist

ance

can

still

bebe

yond

abo

und.

Topr

even

tth

epr

oton

sfr

ombe

ing

pulle

dto

ocl

ose

toge

ther

am

ixed

appr

oach

can

beus

ed.

Inth

isap

proa

chth

epe

nalty

isze

row

hen

the

inst

anta

neou

sdi

stan

ceis

with

inth

ebo

unds

,ot

herw

ise

the

viol

atio

nis

the

squa

rero

otof

the

prod

uct

ofth

ein

stan

tane

ous

viol

atio

nan

dth

etim

eav

erag

edvi

olat

ion.

Page 84: CS USER AL Simulations - GROMACS

64

Ch

ap

ter

4.

Force

field

s

Averaging

overm

ultiplepairs

Som

etimes

itis

unclearfrom

experimentaldata

which

atompair

givesrise

toa

singleN

OE

,in

otheroccasions

itcan

beobvious

thatm

orethan

onepair

contributesdue

tothe

symm

etryof

thesystem

,e.g.

am

ethylgroupw

iththree

protons.F

orsuch

agroup

itis

notpossible

todistinguish

between

theprotons,therefore

theyshould

allbetaken

intoaccountw

hencalculating

thedistance

between

thism

ethylgroupand

anotherproton

(orgroup

ofprotons).D

ueto

thephysicalnature

ofm

agneticresonance,the

intensityofthe

NO

Esignalis

proportionaltothe

distancebetw

eenatom

sto

thepow

erof

-6.T

hus,w

hencom

biningatom

pairs,a

fixedlist

ofN

restraintsm

aybe

takentogether,w

herethe

apparent“distance”is

givenby:

rN

(t)= [

N∑n=

1

rn (t) −

6 ]−

1/6

(4.66)

where

we

userij

oreqn.4.64

forthe

rn .

TherN

oftheinstantaneous

andtim

e-averageddistances

canbe

combined

todo

am

ixedrestraining

asindicated

above.A

sm

orepairs

ofprotonscontribute

tothe

same

NO

Esignal,

theintensity

willincrease,

andthe

summ

ed“distance”

willbe

shorterthan

anyofits

components

dueto

thereciprocalsum

mation.

There

aretw

ooptions

fordistributing

theforces

overthe

atompairs.

Inthe

conservativeoption

theforce

isdefined

asthe

derivateof

therestraint

potentialwith

respectto

thecoordinates.

This

resultsin

aconservative

potentialwhen

notim

eaveraging

isused.

The

forcedistribution

overthe

pairsis

proportionaltor −6.

This

means

thata

closepair

feelsa

much

largerforce

thana

distantpair,

which

might

leadto

a’too

rigid’molecule.

The

otheroption

isan

equalforcedistribution.

Inthis

caseeach

pairfeels1/N

ofthederivative

oftherestraintpotentialw

ithrespectto

rN

.T

headvantage

ofthism

ethodis

thatmore

conformations

mightbe

sampled,butthe

non-conservativenature

oftheforces

canlead

tolocalheating

oftheprotons.

Itis

alsopossible

touseen

sem

ble

ave

ragin

gusingm

ultiple(protein)

molecules.

Inthis

casethe

boundsshould

below

eredas

in:

r1

=r1 ∗

M−

1/6

r2

=r2 ∗

M−

1/6

(4.67)

where

Mis

thenum

berof

molecules.

The

GR

OM

AC

Spreprocessor

gro

mp

pcan

dothis

auto-m

aticallyw

henthe

appropriateoption

isgiven.

The

resulting“distance”

isthen

usedto

calculatethe

scalarforce

accordingto:

Fi

=0

rN<r1

=−k

dr (r

N−r1 ) r

ij

rij

r1≤rN<r2

=−k

dr (r

2 −r1 ) r

ij

rij

rN≥r2

(4.68)

where

iand

jdenote

theatom

sofallthe

pairsthatcontribute

tothe

NO

Esignal.

Using

distancerestraints

Alist

ofdistance

restrainsbased

onN

OE

datacan

beadded

toa

molecule

definitionin

yourtopology

file,likein

thefollow

ingexam

ple:

B.4

.Ta

bula

ted

fun

ction

s1

65

For(127

−E

)=odd

equation(eqn.B.45)

canbe

rewritten

as

y(x)=

(2127−

E−

12

)( 1.F2) −

1/2

(B.52)

thus

E′=

126−E

2+

127(B

.53)

which

alsocan

becalculated

exactlyin

integerarithm

etic.N

otethatthe

fractionis

automatically

correctedfor

itsrange

earlierm

entioned,sothe

exponentdoesnotneed

anextra

correction.

The

conclusionsfrom

thisare:

•T

hefraction

andexponent

lookuptable

areindependent.

The

fractionlookup

tableexists

oftwo

tables(odd

andeven

exponent)so

theodd/even

information

oftheexponent(lsb

bit)has

tobe

usedto

selecttherighttable.

•T

heexponenttable

isan

256x

8bittable,initialized

forodd

andeven

.

B.3.6

Implem

entation

The

lookuptables

canbe

generatedby

asm

allC

program,

which

usesfloating

pointnum

bersand

operationsw

ithIE

EE

32bit

singleprecision

format.

Note

thatbecause

ofthe

odd/ev

eninform

ationthatis

needed,thefraction

tableis

twice

thesize

earlierspecified

(13biti.s.o.

12bit).

The

functionaccording

toequation

(eqn.B

.29)has

tobe

implem

ented.A

ppliedto

the1/ √x

function,equation(eqn.B.28)

leadsto

f=a−

1y2

(B.54)

andso

f′=

2y3

(B.55)

so

yn+

1=y

n−a−

1y2n

2y3n

(B.56)

ory

n+

1=y

n2(3−ay

2n )(B

.57)

Where

y0

canbe

foundin

thelookup

tables,andy1

givesthe

resulttothe

maxim

umaccuracy.

Itis

clearthatonly

oneiteration

extra(in

doubleprecision)

isneeded

fora

doubleprecision

result.

B.4

Tabulatedfunctions

Insom

eofthe

innerloops

ofGR

OM

AC

Slookup

tablesare

usedfor

computation

ofpotentialandforces.

The

tablesare

interpolatedusing

acubic

splinealgorithm

.T

hereare

separatetables

for

Page 85: CS USER AL Simulations - GROMACS

16

4A

pp

en

dix

B.

So

me

imp

lem

en

tatio

nd

eta

ils

B.3

.5S

epar

ate

expo

nent

and

frac

tion

com

puta

tion

The

used

IEE

E32

bits

ingl

epr

ecis

ion

float

ing

poin

tfor

mat

spec

ifies

that

anu

mbe

ris

repr

esen

ted

bya

expo

nent

and

afr

actio

n.T

hepr

evio

usse

ctio

nsp

ecifi

esfo

rev

ery

poss

ible

float

ing

poin

tnu

mbe

rth

elo

okup

tabl

ele

ngth

and

wid

th.

Onl

yth

esi

zeof

the

frac

tion

ofa

float

ing

poin

tnum

ber

defin

esth

eac

cura

cy.

The

conc

lusi

onfr

omth

isca

nbe

that

the

size

ofth

elo

okup

tabl

eis

leng

thof

look

upta

ble,

earli

ersp

ecifi

ed,

times

the

size

ofth

eex

pone

nt(

21228,1Mb)

.T

he1/√x

func

tion

has

the

prop

erty

that

the

expo

nent

isin

depe

nden

toft

hefr

actio

n.T

his

beco

mes

clea

rift

heflo

atin

gpo

intr

epre

sent

atio

nis

used

.D

efine x≡

(−1)

S(2

E−

127)(

1.F

)(B

.42)

see

Fig

.B.1

whe

re0≤S≤

1,0≤E≤

255,

1≤

1.F<

2an

dS

,E,F

inte

ger

(nor

mal

izat

ion

cond

ition

s).

The

sign

bit(S

)ca

nbe

omitt

edbe

caus

e1/√x

ison

lyde

fined

forx>

0.T

he1/√x

func

tion

appl

ied

tox

resu

ltsin

y(x

)=

1 √x

(B.4

3)

or

y(x

)=

1√ (2

E−

127)(

1.F

)(B

.44)

this

can

bere

writ

ten

asy(x

)=

(2E−

127)−

1/2(1.F

)−1/2

(B.4

5)

Defi

ne(2

E′ −

127)≡

(2E−

127)−

1/2

(B.4

6)

1.F′≡

(1.F

)−1/2

(B.4

7)

then

1 √2<

1.F′≤

1ho

lds,

soth

eco

nditi

on1≤

1.F′<

2w

hich

ises

sent

ialf

orno

rmal

ized

real

repr

esen

tatio

nis

not

valid

anym

ore.

By

intr

oduc

ing

anex

tra

term

this

can

beco

rrec

ted.

Rew

rite

the1/√x

func

tion

appl

ied

toflo

atin

gpo

intn

umbe

rs,e

quat

ion

(eqn

.B

.45)

as

y(x

)=

(2127−

E2

−1)(

2(1.F

)−1/2)

(B.4

8)

and

(2E′ −

127)≡

(2127−

E2

−1)

(B.4

9)

1.F′≡

2(1.F

)−1/2

(B.5

0)

then√

2<

1.F≤

2ho

lds.

Thi

sis

not

the

exac

tva

lidra

nge

asde

fined

for

norm

aliz

edflo

atin

gpo

int

num

bers

ineq

uatio

n(e

qn.

B.4

2).

The

valu

e2ca

uses

the

prob

lem

.B

ym

appi

ngth

isva

lue

onth

ene

ares

tre

pres

enta

tion<

2th

isca

nbe

solv

ed.

The

smal

lerr

orth

atis

intr

oduc

edby

this

appr

oxim

atio

nis

with

inth

eal

low

able

rang

e.

The

inte

gerr

epre

sent

atio

nof

the

expo

nent

isth

ene

xtpr

oble

m.

Cal

cula

ting

(2127−

E2

−1)

intr

oduc

esa

frac

tiona

lre

sult

if(12

7−E

)=odd.

Thi

sis

agai

nea

sily

acco

unte

dfo

rby

split

ting

upth

eca

lcul

atio

nin

toan

odd

and

anev

enpa

rt.

For(127−E

)=even

E′in

equa

tion

(eqn

.B.4

9)ca

nbe

exac

tlyca

lcul

ated

inin

tege

rar

ithm

etic

asa

func

tion

ofE

.

E′=

127−E

2+

126

(B.5

1)

4.2

.B

on

de

din

tera

ctio

ns

65

[d

ista

nce

_re

stra

ints

];

ai

aj

typ

ein

de

xty

pe

’lo

wu

p1

up

2fa

c1

01

61

01

0.0

0.3

0.4

1.0

10

28

11

10

.00

.30

.41

.01

04

61

11

0.0

0.3

0.4

1.0

16

22

12

10

.00

.30

.42

.51

63

41

31

0.0

0.5

0.6

1.0

Inth

isex

ampl

ea

num

ber

offe

atur

esca

nbe

foun

d.In

colu

mns

ai

and

aj

you

find

the

atom

num

bers

ofth

epa

rtic

les

tobe

rest

rain

ed.

Thetyp

eco

lum

nsh

ould

alw

ays

be1.

As

expl

aine

din

sec.

4.2.

11,m

ultip

ledi

stan

ces

can

cont

ribut

eto

asi

ngle

NO

Esi

gnal

.In

the

topo

logy

this

can

bese

tus

ing

thein

de

xco

lum

n.In

our

exam

ple,

the

rest

rain

ts10

-28

and

10-4

6bo

thha

vein

dex

1,th

eref

ore

they

are

trea

ted

sim

ulta

neou

sly.

An

extr

are

quire

men

tfor

trea

ting

rest

rain

tsto

geth

er,

isth

atth

ere

stra

ints

shou

ldbe

onsu

cces

sive

lines

,w

ithou

tan

yot

her

inte

rven

ing

rest

rain

t.T

hety

pe

’co

lum

nw

illus

ually

be1,

butc

anbe

sett

o2

toob

tain

adi

stan

cere

stra

intw

hich

will

neve

rbe

time

and

ense

mbl

eav

erag

ed,

this

can

beus

eful

for

rest

rain

ing

hydr

ogen

bond

s.T

heco

lum

nslo

w,u

p1

and

up

2ho

ldth

eva

lues

ofr0,r

1an

dr 2

from

eqn.

4.61

.In

som

eca

ses

itca

nbe

usef

ulto

have

diffe

rent

forc

eco

nsta

nts

for

som

ere

stra

ints

,th

isis

cont

rolle

dby

the

colu

mn

fac

.T

hefo

rce

cons

tant

inth

epa

ram

eter

file

ism

ultip

lied

byth

eva

lue

inth

eco

lum

nfa

cfo

rea

chre

stra

int.

Som

epa

ram

eter

sfo

rN

MR

refin

emen

tcan

besp

ecifi

edin

the

gro

mp

p.m

dp

file:

dis

re:

type

ofdi

stan

cere

stra

inin

g.T

hed

isre

varia

ble

sets

the

type

ofdi

stan

cere

stra

in-

ing.

no

/sim

ple

turn

sth

edi

stan

cere

stra

inin

gof

f/on.

Whe

nm

ultip

lepr

otei

nsor

pep-

tides

are

used

inth

esi

mul

atio

nen

sem

ble

aver

agin

gca

nbe

turn

edon

byse

tting

dis

re=

en

sem

ble

.

dis

rew

eig

htin

g:

forc

e-w

eigh

ting

inre

stra

ints

with

mul

tiple

pairs

.B

yde

faul

t,th

efo

rce

due

toth

edi

stan

cere

stra

inti

sdi

strib

uted

equa

llyov

eral

lthe

pairs

invo

lved

inth

ere

stra

int.

Thi

sca

nal

sobe

expl

icitl

yse

lect

edw

ithdis

rew

eig

htin

g=

eq

ua

l.

Ifyo

uin

stea

dse

tth

isop

tion

tod

isre

we

igh

ting

=co

nse

rva

tive

you

get

cons

erva

tive

forc

esw

hen

dis

reta

u=

0.

dis

rem

ixe

d:

how

toca

lcul

ate

the

viol

atio

ns.d

isre

mix

ed

=n

ogi

ves

norm

altim

eav

-er

aged

viol

atio

ns.

Whe

ndisr

em

ixe

d=

yes

the

squa

rero

otof

the

prod

ucto

fthe

time

aver

aged

and

the

inst

anta

neou

svi

olat

ions

isus

ed.

dis

refc

:fo

rce

cons

tantk

dr

for

dist

ance

rest

rain

ts.k

dr

(eqn

.4.6

1)ca

nbe

set

asva

riabl

ed

isre

fc=

10

00

for

afo

rce

cons

tant

of10

00kJ

mol−1

nm−

2.

Thi

sva

lue

ism

ulti-

plie

dby

the

valu

ein

thefa

cco

lum

nin

the

dist

ance

rest

rain

tent

ries

inth

eto

polo

gyfil

e.

dis

reta

u:

time

cons

tant

for

rest

rain

ts.τ

(eqn

.4.6

5)ca

nbe

set

asva

riabl

edisr

eta

u=

10

for

atim

eco

nsta

ntof

10ps

.T

ime

aver

agin

gca

nbe

turn

edof

fby

setti

ngd

isre

tau

to0.

nst

dis

reo

ut

:pa

irdi

stan

ceou

tput

freq

uenc

y.D

eter

min

esho

wof

ten

the

time

aver

aged

and

inst

anta

neou

sdi

stan

ces

ofal

lat

ompa

irsin

volv

edin

dist

ance

rest

rain

tsar

ew

ritte

nto

the

ener

gyfil

e.

Page 86: CS USER AL Simulations - GROMACS

66

Ch

ap

ter

4.

Force

field

s

4.3F

reeenergy

interactions

This

sectiondescribes

theλ-dependenceof

thepotentials

usedfor

freeenergy

calculations(see

sec.3.12).A

llcomm

ontypes

ofpotentialsand

constraintscan

beinterpolated

smoothly

fromstate

A(λ

=0)

tostate

B(λ

=1)

andvice

versa.A

llbondedinteractions

areinterpolated

bylinear

interpolationof

theinteraction

parameters.

Non-bonded

interactionscan

beinterpolated

linearlyor

viasoft-core

interactions.

Harm

onicpotentials

The

example

givenhere

isfor

thebond

potentialw

hichis

harmonic

inG

RO

MA

CS

.H

owever,

theseequations

applyto

theangle

potentialandthe

improper

dihedralpotentialasw

ell.

Vb

=12((1

−λ)k

Ab+λk

Bb)(b−

(1−λ)b

A0−λbB0

)2

(4.69)

∂V

b

∂λ

=12(k

Bb−k

Ab) [b−

(1−λ)b

A0+λbB0

)2+

(bA0−bB0

)(b−(1−λ)b

A0−λbB0

) ](4.70)

GR

OM

OS

-96bonds

andangles

Fourth

power

bondstretching

andcosine

basedangle

potentialsare

interpolatedby

linearinterpo-

lationofthe

forceconstantand

theequilibrium

position.F

ormulas

arenotgiven

here.

Proper

dihedrals

For

theproper

dihedrals,theequations

aresom

ewhatm

orecom

plicated:

Vd

=((1

−λ)k

Ad+λk

Bd)(1

+cos(n

φφ−

((1−λ)φ

A0+λφ

B0))

(4.71)∂V

d

∂λ

=(k

Bd−k

Ad) [1

+cos(n

φφ−

[(1−λ)φ

A0+λφ

B0])−

((1−λ)k

Ad+λk

Bd)(φ

A0−φ

B0)sin(n

φφ−

[(1−λ)φ

A0+λφ

B0] ]

(4.72)

Note:

thatthem

ultiplicitynφ

cannotbe

parameterized

becausethe

functionshould

remain

peri-odic

onthe

interval[0,2π].

Coulom

binteraction

The

Coulom

binteraction

between

two

particlesofw

hichthe

chargevaries

with

λis:

Vc

=f

εrf r

ij [((1−λ)q

Ai+λqBi

)·((1−λ)q

Aj+λqBi

) ](4.73)

∂V

c

∂λ

=f

εrf r

ij [(qBj−qAj)((1

−λ)q

Ai+λqBi

)+

(qBi−qAi)((1

−λ)q

Aj+λqBj

) ](4.74)

where

f=

14πε0

=138.935

485(see

chapter2)

B.3

.C

om

pu

tatio

no

fthe

1.0

/sqrtfu

nctio

n.

16

3

B.3.4

Specification

ofthelookup

table

Tocalculate

thefunction1/ √

xusing

thepreviously

mentioned

iterationschem

e,itisclearthatthe

firstestim

ationof

thesolution

must

beaccurate

enoughto

getprecise

results.T

herequirem

entsfor

thecalculation

are

•M

aximum

possibleaccuracy

with

theused

IEE

Eform

at

•U

seonly

oneiteration

stepfor

maxim

umspeed

The

firstrequirem

entstates

thatthe

resultof

1/ √

xm

ayhave

arelative

errorεrequal

tothe

εr

ofa

IEE

E32

bitsingle

precisionfloating

pointnum

ber.F

romthis

the1/ √

xof

theinitial

approximation

canbe

derived,rew

ritingthe

definitionof

therelative

errorfor

succeedingsteps,

equation(eqn.B.34)

εny

= √εrn+

1

2f′

yf′′

(B.35)

So

forthe

lookuptable

theneeded

accuracyis

∆YY

= √232−

24

(B.36)

which

definesthe

width

ofthetable

thatmustbe≥

13bit.

Atthis

pointtherelative

errorεrn

ofthelookup

tableis

known.

From

thisthe

maxim

umrelative

errorin

theargum

entcanbe

calculatedas

follows.

The

absoluteerror

∆x

isdefined

as

∆x≡

∆Y

Y′

(B.37)

andthus

∆x

Y=

∆YY

(Y′) −

1(B

.38)

andthus

∆x

=con

stantYY′

(B.39)

forthe

1/ √

xfunction

Y/Y′∼

xholds,

so∆x/x

=con

stant.

This

isa

propertyof

theused

floatingpoint

representationas

earlierm

entioned.T

heneeded

accuracyof

theargum

entof

thelookup

tablefollow

sfrom

∆xx

=−

2∆YY

(B.40)

so,usingthe

floatingpointaccuracy,equation

(eqn.B

.36)

∆xx

=−

2 √232−

24

(B.41)

This

definesthe

lengthofthe

lookuptable

which

shouldbe

≥12

bit.

Page 87: CS USER AL Simulations - GROMACS

16

2A

pp

en

dix

B.

So

me

imp

lem

en

tatio

nd

eta

ils

︸︷︷

︸︸

︷︷︸

?F

ES

023

31

Value

=(−

1)S(2

E−

127)(

1.F

)

023

31

Value

=(−

1)S(2

E−

127)(

1.F

)

Fig

ure

B.1

:IE

EE

sing

lepr

ecis

ion

float

ing

poin

tfor

mat

can

now

beso

lved

usin

gN

ewto

n-R

aphs

on.

An

itera

tion

ispe

rfor

med

byca

lcul

atin

g

y n+

1=y n−f(y

n)

f′ (y n

)(B

.29)

The

abso

lute

erro

rε,in

this

appr

oxim

atio

nis

defin

edby

ε≡y n−q

(B.3

0)

usin

gTa

ylor

serie

sex

pans

ion

toes

timat

eth

eer

ror

resu

ltsin

ε n+

1=−ε2 n 2f′′ (y n

)f′ (y n

)(B

.31)

acco

rdin

gto

[76]e

quat

ion

(3.2

).T

his

isan

estim

atio

nof

the

abso

lute

erro

r.

B.3

.3A

pplie

dto

float

ing

poin

tnum

bers

Flo

atin

gpo

intn

umbe

rsin

IEE

E32

bits

ingl

epr

ecis

ion

form

atha

vea

near

lyco

nsta

ntre

lativ

eer

ror

of∆x/x

=2−

24.

As

seen

earli

erin

the

Tayl

orse

ries

expa

nsio

neq

uatio

n(e

qn.

B.3

1),t

heer

ror

inev

ery

itera

tion

step

isab

solu

tean

din

gene

rald

epen

dent

ofy.

Ifth

eer

ror

isex

pres

sed

asa

rela

tive

erro

rεr

the

follo

win

gho

lds

ε rn+

1≡ε n

+1

y(B

.32)

and

so

ε rn+

1=−

(εn y)2yf′′

2f′

(B.3

3)

for

the

func

tionf

(y)

=y−

2th

ete

rmyf′′ /

2f′ i

sco

nsta

nt(e

qual

to−3/

2)so

the

rela

tive

erro

rε rn

isin

depe

nden

tofy.

ε rn+

1=

3 2(ε

r n)2

(B.3

4)

The

conc

lusi

onof

this

isth

atth

efu

nctio

n1/√x

can

beca

lcul

ated

with

asp

ecifi

edac

cura

cy.

4.3

.F

ree

en

erg

yin

tera

ctio

ns

67

Cou

lom

bin

tera

ctio

nw

ithR

eact

ion

Fie

ld

The

coul

omb

inte

ract

ion

incl

udin

ga

reac

tion

field

,be

twee

ntw

opa

rtic

les

ofw

hich

the

char

geva

ries

withλ

is:

Vc

=f

[ 1 r ij

+k

rfr2 ij

−c r

f

] [ ((1−λ)q

A i+λqB i

)·(

(1−λ)q

A j+λqB i

)](4

.75)

∂V

c

∂λ

=f

[ 1 r ij

+k

rfr2 ij

−c r

f

] ·[ (q

B j−qA j

)((1−λ)q

A i+λqB i

)+

(qB i−qA i

)((1−λ)q

A j+λqB j

)](4

.76)

Not

eth

atth

eco

nsta

ntsk r

fan

dc r

far

ede

fined

usin

gth

edi

elec

tric

cons

tantε rf

ofth

em

ediu

m(s

eese

c.4.1

.4).

Lenn

ard-

Jone

sin

tera

ctio

n

For

the

Lenn

ard-

Jone

sin

tera

ctio

nbe

twee

ntw

opa

rtic

les

ofw

hich

the

ato

mty

pev

arie

sw

ithλ

we

can

writ

e:

VL

J=

((1−λ)C

A 12+λC

B 12)

r12

ij

−(1−λ)C

A 6+λC

B 6

r6 ij

(4.7

7)

∂V

LJ

∂λ

=C

B 12−C

A 12

r12

ij

−C

B 6−C

A 6

r6 ij

(4.7

8)

Itsh

ould

beno

ted

that

itis

also

poss

ible

toex

pres

sa

path

way

from

stat

eA

tost

ate

Bus

ing

σan

(see

eqn.4

.5).

Itm

ayse

emto

mak

ese

nse

phys

ical

ly,

tova

ryth

efo

rcefi

eld

para

met

ers

σan

rath

erth

anth

ede

rived

para

met

ers

C12

andC

6.

How

ever

,the

diffe

renc

ebe

twee

nth

epa

thw

ays

inpa

ram

eter

spac

eis

notl

arge

,and

the

free

ener

gyits

elfd

oes

notd

epen

don

the

path

way

,the

refo

rew

eus

eth

esi

mpl

efo

rmul

atio

npr

esen

ted

abov

e.

Kin

etic

Ene

rgy

Whe

nth

em

ass

ofa

part

icle

chan

ges

ther

eis

also

aco

ntrib

utio

nof

the

kine

ticen

ergy

toth

efr

eeen

ergy

(not

eth

atw

eca

nno

tw

rite

the

mom

entu

mp

asmv

sinc

eth

atw

ould

resu

ltin

the

sign

of∂E

k∂λ

bein

gin

corr

ect[5

2]):

Ek

=1 2

p2

(1−λ)m

A+λm

B(4

.79)

∂Ek

∂λ

=−

1 2p

2(m

B−m

A)

((1−λ)m

A+λm

B)2

(4.8

0)

afte

rta

king

the

deriv

ativ

e,w

ecan

inse

rtp

=m

v,s

uch

that

:

∂Ek

∂λ

=−

1 2v

2(m

B−m

A)

(4.8

1)

Page 88: CS USER AL Simulations - GROMACS

68

Ch

ap

ter

4.

Force

field

s

0 �

0.51�

1.52 �

2.53 �

r�

−1 0 1 2 3 4 5

Vsc�

LJ, α=

0LJ, α

=1.5

LJ, α=

23/r,

α=

03/r,

α=

1.53/r,

α=

2

Figure

4.14:S

oft-coreinteractions

atλ

=0.5,w

ithC

A6=C

A12

=C

B6=C

B12

=1.

Constraints

The

constraintsare

formally

partoftheH

amiltonian,and

thereforethey

givea

contributionto

thefree

energy.In

GR

OM

AC

Sthis

canbe

calculatedusing

theLIN

CS

orthe

SH

AK

Ealgorithm

.If

we

havea

number

ofconstraintequationsgk :

gk

=rk−d

k(4.82)

where

rk

isthe

distancevector

between

two

particlesandd

kis

theconstraint

distancebetw

eenthe

two

particlesw

ecan

write

thisusing

dependentdistanceas

gk

=rk− ((1

−λ)d

Ak+λd

Bk )(4.83)

thecontributionC

λto

theH

amiltonian

usingLagrange

multipliers

λ:

= ∑k

λk g

k(4.84)

∂C

λ

∂λ

= ∑k

λk (d

Bk−d

Ak )(4.85)

4.3.1S

oft-coreinteractions

The

linearinterpolation

oftheLennard-Jones

andC

oulomb

potentialsgives

problems

when

grow-

ingparticles

outofnothingor

when

making

particlesdisappear

closeto

0or

1).To

circumvent

theseproblem

s,thesingularities

inthe

potentialsneed

tobe

removed.

This

isdone

with

soft-corepotentials.

InG

RO

MA

CS

thesoft-core

potentialV

scis:

Vsc (r)

=(1−λ)V

A(r

A)+λV

B(r

B)

(4.86)

rA

= (ασ

6Aλ

2+r6 )

16(4.87)

B.3

.C

om

pu

tatio

no

fthe

1.0

/sqrtfu

nctio

n.

16

1

Note

thatthis

loopprovides

much

lessoptim

izationthan

thew

aterloop,

butit

isslightly

betterthan

thedefaultroutine.

The

gainof

theseim

plementations

isthat

thereare

more

floatingpoint

operationsin

asingle

loop,w

hichim

pliesthat

some

compilers

canschedule

thecode

better.H

owever,

itturns

outthat

evensom

eofthe

mostadvanced

compilers

haveproblem

sw

ithscheduling,im

plyingthatm

anualtw

eakingis

necessaryto

getoptim

umperform

ance.T

hism

ayinclude

comm

on-subexpressionelim

ination,orm

ovingcode

around.

B.2.2

Fortran

Code

Unfortunately,F

ortrancom

pilersare

stillbetterthan

C-com

pilers,form

ostmachines

anyway.

For

some

machines

(e.g.SG

IPow

erC

hallenge)the

differencem

aybe

upto

afactor

of3,in

thecase

ofvectorcom

putersthis

may

beeven

larger.T

herefore,some

oftheroutines

thattakeup

alotof

computer

time

havebeen

translatedinto

Fortran

andeven

assembly

codefor

IntelandA

MD

x86processors.

Inm

ostcases,the

Fortran

orassem

blyloops

shouldbe

selectedautom

aticallyby

theconfigure

scriptwhen

appropriate,butyoucan

alsotw

eakthis

bysetting

optionsto

theconfigure

script.

B.3

Com

putationofthe

1.0/sqrtfunction.

B.3.1

Introduction.

The

GR

OM

AC

Sprojectstarted

with

thedevelopm

entofa1/ √

xprocessor

which

calculates

Y(x)

=1√x

(B.25)

As

theprojectcontinued,the

Inteli860

processorw

asused

toim

plementG

RO

MA

CS

,which

nowturned

intoalm

ostafullsoftw

areproject.

The1/ √

xprocessor

was

implem

entedusing

aN

ewton-

Raphson

iterationschem

efor

onestep.

For

thisit

neededlookup

tablesto

providethe

initialapproxim

ation.T

he1/ √x

functionm

akesitpossible

touse

two

almostindependenttables

forthe

exponentseedand

thefraction

seedw

iththe

IEE

Efloating

pointrepresentation.

B.3.2

General

According

to[76]

the1/ √

xcan

becalculated

usingthe

New

ton-Raphson

iterationschem

e.T

heinverse

functionis

X(y)

=1y2

(B.26)

So

insteadofcalculating

Y(a)

=q

(B.27)

theequation

X(q)−

a=

0(B

.28)

Page 89: CS USER AL Simulations - GROMACS

16

0A

pp

en

dix

B.

So

me

imp

lem

en

tatio

nd

eta

ils

B.1

.5V

irial

from

Sha

ke.

An

impo

rtan

tcon

trib

utio

nto

the

viria

lcom

esfr

omsh

ake.

Sat

isfy

ing

the

cons

trai

nts

afo

rce

Gis

exer

ted

onth

epa

rtic

les

shak

en.

Ifth

isfo

rce

does

not

com

eou

tof

the

algo

rithm

(as

inst

anda

rdsh

ake)

itca

nbe

calc

ulat

edaf

terw

ards

(whe

nus

ing

lea

p-f

rog)

by:

∆r

i=

ri(t+

∆t)−

[ri(t)

+v

i(t−

∆t 2)∆t+

Fi

mi∆t2

](B

.22)

Gi

=m

i∆r

i

∆t2

(B.2

3)

but

this

does

not

help

usin

the

gene

ralc

ase.

Onl

yw

hen

nope

riodi

city

isne

eded

(like

inrig

idw

ater

)th

isca

nbe

used

,oth

erw

ise

we

mus

tadd

the

viria

lcal

cula

tion

inth

ein

ner

loop

ofsh

ake.

Whe

nit

isap

plic

able

the

viria

lcan

beca

lcul

ated

inth

esi

ngle

sum

way

:

Ξ=

−1 2

Nc ∑ i

ri⊗

Fi

(B.2

4)

whe

reN

cis

the

num

ber

ofco

nstr

aine

dat

oms.

B.2

Opt

imiz

atio

ns

Her

ew

ede

scrib

eso

me

ofth

eal

gorit

hmic

optim

izat

ions

used

inG

RO

MA

CS

,ap

art

from

par-

alle

lism

.O

neof

thes

e,th

eim

plem

enta

tion

ofth

e1.

0/sq

rt(x

)fu

nctio

nis

trea

ted

sepa

rate

lyin

sec.

B.3

.T

hem

osti

mpo

rtan

toth

erop

timiz

atio

nsar

ede

scrib

edbe

low

.

B.2

.1In

ner

Loop

sfo

rW

ater

GR

OM

AC

Sus

ers

spec

iali

nner

loop

toca

lcul

ate

non-

bond

edin

tera

ctio

nsfo

rwat

erm

olec

ules

with

othe

rat

oms,

and

yeta

noth

erse

tofl

oops

for

inte

ract

ions

betw

een

pairs

ofw

ater

mol

ecul

es.

Thi

sve

ryop

timiz

edlo

opas

sum

esa

wat

erm

odel

sim

ilar

toS

PC

[48

],i.e

.:

1.T

here

are

thre

eat

oms

inth

em

olec

ule.

2.T

hefir

stat

omha

sLe

nnar

d-Jo

nes

(sec

.4.

1.1)

and

coul

omb

(sec

.4.1.

3)in

tera

ctio

ns.

3.A

tom

stw

oan

dth

ree

have

only

coul

omb

inte

ract

ions

,and

equa

lcha

rges

.

The

loop

also

wor

ksfo

rth

eS

PC

/E[

75]

and

TIP

3P[42

]w

ater

mod

els.

For

mor

eco

mpl

icat

edm

olec

ules

ther

eis

age

nera

lsol

vent

loop

assu

min

g(n

ote

the

orde

r):

1.A

tth

ebe

ginn

ing

ofth

em

olec

ule

topo

logy

ther

eis

anar

bitr

ary

num

ber

ofat

oms

with

Lenn

ard-

Jone

san

dco

ulom

bin

tera

ctio

ns.

2.T

hen

we

have

anar

bitr

ary

num

ber

ofat

oms

with

coul

omb

inte

ract

ions

only

.

3.A

ndfin

ally

ther

eca

nbe

anar

bitr

ary

num

bero

fato

ms

with

Lenn

ard-

Jone

sin

tera

ctio

nson

ly.

4.4

.M

eth

od

s6

9

i+1

i+3

ii+

2i+

4

Fig

ure

4.15

:A

tom

sal

ong

anal

kane

chai

n.

r B=( ασ

6 B(1−λ)2

+r6)1 6

(4.8

8)

whe

reV

Aan

dV

Bar

eth

eno

rmal

’har

dco

re’V

ande

rW

aals

orE

lect

rost

atic

pote

ntia

lsin

stat

eA

(λ=

0)an

dst

ate

B(λ

=1)

resp

ectiv

ely,α

isth

eso

ft-co

repa

ram

eter

,whi

chm

ainl

yco

ntro

lsth

ehe

ight

ofth

epo

tent

iala

roun

dr=

0,σ

isth

era

dius

ofth

ein

tera

ctio

n,w

hich

is(C12/C

6)1

/6

ora

pred

efine

dva

lue

whe

nC6

orC

12

isze

ro.

For

inte

rmed

iateλ

,r A

andr B

alte

rth

ein

tera

ctio

nsve

rylit

tlew

henr>α

1/6σ

and

they

quic

kly

switc

hth

eso

ft-co

rein

tera

ctio

nto

anal

mos

tcon

stan

tva

lue

whe

nrbe

com

essm

alle

r(F

ig.4.14

).T

hefo

rce

is:

Fsc

(r)

=−∂V

sc(r

)∂r

=(1−λ)F

A(r

A)( r r A

) 5 +λF

B(r

B)( r r B

) 5(4

.89)

whe

reF

Aan

dF

Bar

eth

e’h

ard

core

’for

ces.

The

cont

ribut

ion

toth

ede

rivat

ive

ofth

efr

eeen

ergy

is: ∂V

sc(r

)∂λ

=−V

A(r

A)+V

B(r

B)+

1 3αλ(1−λ)( −

FA(r

A)σ

6 Ar−

5A

+F

B(r

B)σ

6 Br−

5B

) (4.9

0)

4.4

Met

hods

4.4.

1E

xclu

sion

san

d1-

4In

tera

ctio

ns.

Ato

ms

with

ina

mol

ecul

eth

atar

ecl

ose

byin

the

chai

n,i.e

.at

oms

that

are

cova

lent

lybo

nded

,or

linke

dby

one

resp

ectiv

ely

two

atom

sar

eso

-cal

led

first

ne

igh

bo

rs,

seco

nd

ne

igh

bo

rsand

third

ne

igh

bo

rs,(s

eeF

ig.4

.15)

.S

ince

the

inte

ract

ions

ofat

omiw

ithi+

1

and

the

inte

ract

ion

ofat

omi

with

atom

i+2

are

mai

nly

quan

tum

mec

hani

cal,

they

can

not

bem

odel

edby

aLe

nnar

d-Jo

nes

pote

ntia

l.In

stea

dit

isas

sum

edth

atth

ese

inte

ract

ions

are

adeq

uate

lym

odel

edby

aha

rmon

icbo

ndte

rmor

cons

trai

nt(

i,i+

1)an

da

harm

onic

angl

ete

rm(

i,i+

2).

The

first

and

seco

ndne

ighb

ors

(ato

ms

i+1a

ndi+

2)ar

eth

eref

oree

xclu

de

dfro

mth

eLe

nnar

d-Jo

nes

inte

ract

ion

listo

fato

mi;a

tom

si+

1an

di+

2ar

eca

llede

xclu

sio

nso

fato

mi.

For

third

neig

hbor

sth

eno

rmal

Lenn

ard-

Jone

sre

puls

ion

isso

met

imes

still

too

stro

ng,w

hich

mea

nsth

atw

hen

appl

ied

toa

mol

ecul

eth

em

olec

ule

wou

ldde

form

orbr

eak

due

toth

ein

tern

alst

rain

.T

his

ises

peci

ally

the

case

for

Car

bon-

Car

bon

inte

ract

ions

ina

cis-

conf

orm

atio

n(e

.g.ci

s-bu

tane

).T

here

fore

for

som

eof

thes

ein

tera

ctio

nsth

eLe

nnar

d-Jo

nes

repu

lsio

nha

sbe

enre

duce

din

the

GR

OM

OS

forc

efie

ld,w

hich

isim

plem

ente

dby

keep

ing

ase

para

telis

tof1

-4an

dno

rmal

Lenn

ard-

Jone

spa

ram

eter

s.In

othe

rfor

cefie

lds,

such

asO

PLS

[50

],th

est

anda

rdLe

nnar

d-Jo

nes

para

met

ers

are

redu

ced

bya

fact

orof

two,

buti

nth

atca

seal

soth

edi

sper

sion

(r−

6)a

ndth

eco

ulom

bin

tera

ctio

nar

esc

aled

.G

RO

MA

CS

can

use

eith

erof

thes

em

etho

ds.

Page 90: CS USER AL Simulations - GROMACS

70

Ch

ap

ter

4.

Force

field

s

4.4.2C

hargeG

roups.

Inprinciple

theforce

calculationin

MD

isanO

(N2)

problem.

Therefore

we

applya

cut-offfor

non-bondedforce

(NB

F)

calculations:only

theparticles

within

acertain

distanceof

eachother

areinteracting.

This

reducesthe

costtoO

(N)

(typically100

Nto

200N

)of

theN

BF.

Italso

introducesan

error,which

is,inm

ostcases,acceptable,exceptwhen

applyingthe

cut-offimplies

thecreation

ofcharges,inw

hichcase

youshould

considerusing

thelattice

summ

ethodsprovided

byG

RO

MA

CS

.

Consider

aw

aterm

oleculeinteracting

with

anotheratom

.W

henw

ew

ouldapply

thecut-offon

anatom

-atombasis

we

might

includethe

atom-O

xygeninteraction

(with

acharge

of-0.82)

without

thecom

pensatingcharge

oftheH

ydrogensand

soinduce

alarge

dipolem

omentover

thesystem

.T

hereforew

ehave

tokeep

groupsof

atoms

with

totalcharge

0together,

theso-called

cha

rgeg

rou

ps.

4.4.3Treatm

entofcut-offs

GR

OM

AC

Sis

quiteflexible

intreating

cut-offs,w

hichim

pliesthere

canbe

quitea

number

ofparam

etersto

set.T

heseparam

etersare

setin

theinput

filefor

grompp.

There

aretw

osort

ofparam

etersthat

affectthe

cut-offinteractions;

youcan

selectw

hichtype

ofinteraction

touse

ineach

case,andw

hichcut-offs

shouldbe

usedin

theneighborsearching.

For

bothC

oulomb

andvan

derW

aalsinteractions

thereare

interactiontype

selectors(term

edvd

wtyp

eand

cou

lom

btyp

e)

andtw

oparam

eters,for

atotal

ofsix

nonbondedinteraction

parameters.

See

sec.7.3.1

fora

complete

descriptionofthese

parameters.

The

neighborsearching

(NS

)can

beperform

edusing

asingle-range,

ora

twin-range

approach.S

incethe

former

ism

erelya

specialcaseofthe

latterw

ew

illdiscussthe

more

generaltwin-range.

Inthis

caseN

Sis

describedby

two

radiirlist

andm

ax(rcou

lom

b,rvd

w).

Usually

onebuilds

theneighbor

listevery10

time

stepsor

every20

fs(param

etern

stlist).

Inthe

neighborlistall

interactionpairs

thatfallw

ithinrlistare

stored.F

urthermore,

theinteractions

between

pairsthatdo

notfallwithinrlist

butdofallw

ithinand

max(rco

ulo

mb

,rvdw

)arecom

putedduring

NS

,andthe

forcesand

energyare

storedseparately,and

addedto

short-rangeforces

ateverytim

estep

between

successiveN

S.Ifrlist

=m

ax(rcou

lom

b,rvd

w)

noforces

areevaluated

duringneighbor

listgeneration.T

hevirialis

calculatedfrom

thesum

oftheshort-

andlong-range

forces.T

hism

eansthat

thevirialcan

beslightly

asymm

etricalatnon-N

Ssteps.

Insingle

precisionthe

virialisalm

ostalways

asymm

etrical,becausethe

off-diagonalelements

areaboutas

largeas

eachelem

entin

thesum

.In

most

casesthis

isnot

reallya

problem,

sincethe

fluctuationsin

devirial

canbe

2orders

ofmagnitude

largerthan

theaverage.

Except

forthe

plaincut-off,

allof

theinteraction

functionsin

Table4.2

requirethat

neighborsearching

isdone

with

alarger

radiusthan

therc

specifiedfor

thefunctionalform

,becauseofthe

useof

chargegroups.

The

extraradius

istypically

ofthe

orderof

0.25nm

(roughlythe

largestdistance

between

two

atoms

ina

chargegroup

plusthe

distancea

chargegroup

candiffuse

within

neighborlistupdates).

B.1

.S

ing

leS

um

Viria

linG

RO

MA

CS

.1

59

The

algorithmto

generatesuch

alistcan

bederived

fromgraph

theory,consideringeach

particlein

am

oleculeas

abead

ina

graph,thebonds

asedges.

1representthe

bondsand

atoms

asbidirectionalgraph

2m

akeallatom

sw

hite

3m

akeone

ofthew

hiteatom

sblack

(atomi)andputitin

thecentralbox

4m

akeallofthe

neighborsofithatare

currentlyw

hite,grey

5pick

oneofthe

greyatom

s(atomj),give

itthecorrectperiodicity

with

respecttoany

ofitsblack

neighborsand

make

itblack

6m

akeallofthe

neighborsofjthatare

currentlyw

hite,grey

7ifany

greyatom

remains,goto

[5]

8ifany

white

atomrem

ains,goto[3]

Using

thisalgorithm

we

can

•optim

izethe

bondedforce

calculationas

wellas

shake

•calculate

thevirialfrom

thebonded

forcesin

thesingle

sumw

ayagain

Find

arepresentation

ofthebonds

asa

bidirectionalgraph.

B.1.4

Virialfrom

CovalentB

onds.

The

covalentbondforce

givesa

contributionto

thevirial,w

ehave

b=

‖r

nij ‖(B

.15)

Vb

=12k

b (b−b0 )

2(B

.16)

Fi

=−∇V

b(B

.17)

=k

b (b−b0 ) r

nij

b(B

.18)

Fj

=−

Fi

(B.19)

The

virialcontributionfrom

thebonds

thenis

Ξb

=−

12(r

ni⊗

Fi

+r

j ⊗F

j )(B

.20)

=−

12r

nij ⊗F

i(B

.21)

Page 91: CS USER AL Simulations - GROMACS

15

8A

pp

en

dix

B.

So

me

imp

lem

en

tatio

nd

eta

ils

Ina

tric

linic

syst

emth

ere

are

27po

ssib

leim

ages

ofi,

whe

ntr

unca

ted

octa

hedr

onis

used

ther

ear

e15

poss

ible

imag

es.

B.1

.2V

irial

from

non-

bond

edfo

rces

.

Her

eth

ede

rivat

ion

for

the

sing

lesu

mvi

riali

nth

en

on

-bo

nd

ed

forc

erout

ine

isgi

ven.i6=j

inal

lfo

rmul

aebe

low

.

Ξ=

−1 2

N ∑ i<j

rn ij⊗

Fij

(B.5

)

=−

1 4

N ∑ i=1

N ∑ j=1

(ri+δ i−

rj)⊗

Fij

(B.6

)

=−

1 4

N ∑ i=1

N ∑ j=1

(ri+δ i

)⊗

Fij−

rj⊗

Fij

(B.7

)

=−

1 4

N ∑ i=1

N ∑ j=1

(ri+δ i

)⊗

Fij−

N ∑ i=1

N ∑ j=1

rj⊗

Fij

(B

.8)

=−

1 4

N ∑ i=1

(ri+δ i

)⊗

N ∑ j=1

Fij−

N ∑ j=1

rj⊗

N ∑ i=1

Fij

(B

.9)

=−

1 4

N ∑ i=1

(ri+δ i

)⊗

Fi

+N ∑ j=

1

rj⊗

Fj

(B

.10)

=−

1 4

( 2N ∑ i=

1

ri⊗

Fi+

N ∑ i=1

δ i⊗

Fi)

(B.1

1)

Inth

ese

form

ulae

we

intr

oduc

ed

Fi

=N ∑ j=

1

Fij

(B.1

2)

Fj

=N ∑ i=

1

Fji

(B.1

3)

whi

chis

the

tota

lfor

ceoni

resp

.j.

Bec

ause

we

use

New

ton’

sth

irdla

w

Fij

=−

Fji

(B.1

4)

we

mus

tin

the

impl

emen

tatio

ndo

uble

the

term

cont

aini

ngth

esh

iftδ i

.

B.1

.3T

hein

tram

olec

ular

shift

(mol

-shi

ft).

For

the

bond

ed-f

orce

san

dsh

ake

itis

poss

ible

tom

ake

am

ol-sh

iftlis

t,in

whi

chth

epe

riodi

city

isst

ored

.W

esi

mpl

eha

vean

arra

ym

shift

inw

hich

for

each

atom

anin

dex

inth

eshift

vec

arra

yis

stor

ed.

4.5

.D

um

my

ato

ms.

71

Type

Par

amet

ers

Cou

lom

bP

lain

cut-

off

r c,ε

r

Rea

ctio

nfie

ldr c

,εrf

Shi

ftfu

nctio

nr 1

,rc,ε

r

Sw

itch

func

tion

r 1,r

c,ε

r

VdW

Pla

incu

t-of

fr c

Shi

ftfu

nctio

nr 1

,rc

Sw

itch

func

tion

r 1,r

c

Tabl

e4.

2:P

aram

eter

sfo

rth

edi

ffere

ntfu

nctio

nalf

orm

sof

the

non-

bond

edin

tera

ctio

ns.

4.5

Dum

my

atom

s.

Dum

my

atom

sca

nbe

used

inG

RO

MA

CS

ina

num

ber

ofw

ays.

We

writ

eth

epo

sitio

nof

the

dum

my

part

icle

rd

asa

func

tion

ofth

epo

sitio

nsof

othe

rpa

rtic

les

ri:

rd

=f(r

1..r

n).

The

dum

my,

whi

chm

ayca

rry

char

ge,o

rca

nbe

invo

lved

inot

her

inte

ract

ions

can

now

beus

edin

the

forc

eca

lcul

atio

n.T

hefo

rce

actin

gon

the

dum

my

part

icle

mus

tbe

redi

strib

uted

over

the

atom

sin

aco

nsis

tent

way

.A

good

way

todo

this

can

befo

und

inre

f.[

53].

We

can

writ

eth

epo

tent

ial

ener

gyas

V=V

(rd,r

1..r

n)

=V∗ (

r1..r

n)

(4.9

1)

The

forc

eon

the

part

iclei

isth

en

Fi=−∂V∗

∂r

i=−∂V

∂r

i−∂r

d

∂r

i

∂V

∂r

d=

Fdir

ect

i+

F′ i

(4.9

2)

the

first

term

ofw

hich

isth

eno

rmal

forc

e.T

hese

cond

term

isth

efo

rce

onpa

rtic

lei

due

toth

edu

mm

ypa

rtic

le,w

hich

can

bew

ritte

nin

tens

orno

tatio

n:

F′ i=

∂x

d

∂x

i

∂y d∂x

i

∂z d∂x

i∂x

d

∂y i

∂y d∂y i

∂z d∂y i

∂x

d

∂z i

∂y d∂z i

∂z d∂z i

Fd

(4.9

3)

whe

reF

dis

the

forc

eon

the

dum

my

part

icle

andx

d,y

dan

dz d

are

the

coor

dina

tes

ofth

edu

mm

ypa

rtic

le.

Inth

isw

ayth

eto

talf

orce

and

the

tota

ltor

que

are

cons

erve

d[

53].

As

afu

rthe

rno

te,t

heco

mpu

tatio

nof

the

viria

l(eq

n.3.18

)vi

riali

sno

n-tr

ivia

lwhe

ndu

mm

yat

oms

are

used

.S

ince

the

viria

linv

olve

sa

sum

mat

ion

over

allt

heat

oms

(rat

hert

han

virt

ualp

artic

les)

the

forc

esm

ostb

ere

dist

ribut

edfr

omth

edu

mm

ies

toth

eat

oms

(usi

ngeq

n.4.

93)

be

fore

com

puta

tion

ofth

evi

rial.

Inso

me

spec

ial

case

sw

here

the

forc

eson

the

atom

sca

nbe

writ

ten

asa

linea

rco

mbi

natio

nof

the

forc

eson

the

dum

mie

s(t

ypes

2an

d3

belo

w)

ther

eis

nodi

ffere

nce

betw

een

com

putin

gth

evi

rial

befo

rean

daf

ter

the

redi

strib

utio

nof

forc

es.

How

ever

,in

the

gene

ralc

ase

redi

strib

utio

nsh

ould

bedo

nefir

st.

The

rear

esi

xw

ays

toco

nstr

uctd

umm

ies

from

surr

ound

ing

atom

sin

GR

OM

AC

S,

whi

chw

eca

t-eg

oriz

eba

sed

onth

enu

mbe

rof

cons

truc

ting

atom

s.N

ote

that

alld

umm

ies

type

sm

entio

ned

can

Page 92: CS USER AL Simulations - GROMACS

72

Ch

ap

ter

4.

Force

field

s

����� ���������� �������������������������� ���������� ����� | |

3fd

| || |

1-aab

a

1-a

a

��������������

����������

������

����

23fad

3ou

t4fd c

b

3

��������������

����������

θ

d

� � � �

����������

������������������� �����

������������������� �����

������������������� �����

��������������

����������

����������

����������

����������

����������

��������������

����������

����������

����������

Figure

4.16:T

hesix

differenttypesofdum

my

atomconstruction

inG

RO

MA

CS

,theconstructing

atoms

areshow

nas

blackcircles,the

dumm

yatom

sin

grey.

beconstructed

fromtypes

3fd(norm

alized,in-plane)

and3out

(non-normalized,

outof

plane).H

owever,

theam

ountof

computation

involvedincreases

sharplyalong

thislist,

soit

isstrongly

recomm

endedto

always

usethe

firstdumm

ytype

thatwillbe

sufficientfora

certainpurpose.

An

overviewofthe

dumm

yconstructions

isgiven

inF

ig.4.16.

2.A

sa

linearcom

binationoftw

oatom

s(F

ig.4.16

2):

rd

=r

i +ar

ij(4.94)

inthis

casethe

dumm

yis

onthe

linethrough

atoms

iand

j.T

heforce

onparticlesi

andj

dueto

theforce

onthe

dumm

ycan

becom

putedas:

F′i

=(1−a)F

d

F′j

=a

Fd

(4.95)

3.A

sa

linearcom

binationofthree

atoms

(Fig.4.16

3):

rd

=r

i +ar

ij+br

ik(4.96)

inthis

casethe

dumm

yis

inthe

planeofthe

otherthree

particles.T

heforce

onparticlesi,j

andk

dueto

theforce

onthe

dumm

ycan

becom

putedas:

F′i

=(1−a−b)F

d

F′j

=a

Fd

F′k

=bF

d

(4.97)

3fd.In

theplane

ofthreeatom

s,with

afixed

distance(F

ig.4.16

3fd):

rd

=r

i +b

rij

+ar

jk

|rij

+ar

jk |(4.98)

inthis

casethe

dumm

yis

inthe

planeofthe

otherthree

particlesata

distanceof

|b|fromi.

The

forceon

particlesi,j

andk

dueto

theforce

onthe

dumm

ycan

becom

putedas:

F′i

=F

d −γ(F

d −p)

F′j

=(1−a)γ(F

d −p)

F′k

=aγ(F

d −p)

where

γ=

b

|rij

+ar

jk |

p=

rid ·F

d

rid ·r

idr

id

(4.99)

Appendix

B

Som

eim

plementation

details

Inthis

chapterw

ew

illpresentsom

eim

plementation

details.T

hisis

farfrom

complete,

butw

edeem

editnecessary

toclarify

some

thingsthatw

ouldotherw

isebe

hardto

understand.

B.1

Single

Sum

Virialin

GR

OM

AC

S.

The

virialΞcan

bew

rittenin

fulltensorform

as:

Ξ=

−12

N∑i<j

rij ⊗

Fij

(B.1)

where⊗

denotesthed

irect

pro

du

ctoftw

ovectors 1.

When

thisis

computed

inthe

innerloop

ofan

MD

program9

multiplications

and9

additionsare

needed2.

Here

itisshow

nhow

itispossible

toextractthe

virialcalculationfrom

theinner

loop[

74].

B.1.1

Virial.

Ina

systemw

ithP

eriodicB

oundaryC

onditions,theperiodicity

mustbe

takeninto

accountforthe

virial:

Ξ=

−12

N∑i<j

rnij ⊗

Fij

(B.2)

wherer

nijdenotes

thedistance

vectorofthen

ea

restim

ageofatomi

fromatom

j.In

thisdefinition

we

addash

iftvecto

rδi

tothe

positionvectorr

iofatom

i.T

hedifference

vectorrnij

isthus

equalto:

rnij

=r

i +δi −

rj

(B.3)

orin

shorthand:r

nij=

rni−

rj

(B.4)

1(u⊗

v)α

β=

uαv

β2T

hecalculation

ofLennard-Jonesand

Coulom

bforces

isabout50

floatingpointoperations.

Page 93: CS USER AL Simulations - GROMACS

15

6A

pp

en

dix

A.

Tech

nic

alD

eta

ils

2.IA

MC

OO

L,if

this

isex

plic

itly

sett

oN

Oyo

urG

RO

MA

CS

life

will

bedu

llan

dbo

ring.

(i.e.

,no

cool

quot

es).

3.W

HE

RE,w

hen

setp

rintd

ebug

ging

info

onlin

enu

mbe

rs.

4.L

OG

BU

FS,

the

size

ofth

ebu

ffer

for

file

I/O.W

hen

sett

o0,

allfi

leI/O

will

beun

buffe

red

and

ther

efor

eve

rysl

ow.

Thi

sca

nbe

hand

yfo

rde

bugg

ing

purp

oses

,bec

ause

iten

sure

sth

atal

lfile

sar

eal

way

sto

tally

up-t

o-da

te.

5.G

MX

NP

RI,f

orS

GIs

yste

ms

only

.W

hen

set,

give

sth

ede

faul

tnon

-deg

radi

ngpr

iorit

y(n

pri)

form

dru

n,n

mru

n,g

cova

ran

dgn

me

ig,e

.g.s

ettin

gse

ten

vG

MX

NP

RI

25

0ca

uses

allr

uns

tobe

perf

orm

edat

near

-low

estp

riorit

yby

defa

ult.

6.G

MXV

IEW

XP

M,G

MXV

IEW

XV

G,G

MXV

IEW

EP

San

dGM

XV

IEW

PD

B,co

mm

ands

used

toau

tom

atic

ally

view

resp

..xvg

,.xp

m,.

ep

san

d.p

db

file

type

s.D

efau

lttox

v,x

mg

r,g

ho

stvi

ew

and

rasm

ol

.S

etto

empt

yto

disa

ble

auto

mat

icvi

ewin

gof

apa

rtic

ular

file

type

.T

heco

mm

and

will

befo

rked

offa

ndru

nin

the

back

grou

ndat

the

sam

epr

iorit

yas

the

GR

OM

AC

Sto

ol(w

hich

mig

htno

tbe

wha

tyo

uw

ant)

.B

eca

refu

lnot

tous

ea

com

man

dw

hich

bloc

ksth

ete

rmin

al(e

.g.vi

),si

nce

mul

tiple

inst

ance

sm

ight

beru

n.

Som

eot

her

envi

ronm

entv

aria

bles

are

spec

ific

toon

epr

ogra

m,s

uch

asT

OTA

Lfo

rth

ed

osh

iftpr

ogra

m,a

ndD

SP

Pfo

rth

edod

ssp

prog

ram

.

4.5

.D

um

my

ato

ms.

73

3fad

.In

the

plan

eof

thre

eat

oms,

with

afix

edan

gle

and

dist

ance

(Fig

.4.

163f

ad):

rd

=r

i+d

cosθ

rij

|rij|+

dsi

r⊥

|r⊥|

whe

rer⊥

=r

jk−

rij·r

jk

rij·r

ijr

ij(4

.100

)

inth

isca

seth

edu

mm

yis

inth

epl

ane

ofth

eot

her

thre

epa

rtic

les

ata

dist

ance

of|d|f

romi

atan

angl

eofα

with

rij

.A

tomk

defin

esth

epl

ane

and

the

dire

ctio

nof

the

angl

e.N

ote

that

inth

isca

seb

andα

mus

tbe

spec

ified

inst

ead

ofaan

db

(see

also

sec.5

.2.2

).T

hefo

rce

onpa

rtic

lesi

,jan

dk

due

toth

efo

rce

onth

edu

mm

yca

nbe

com

pute

das

(with

r⊥

asde

fined

ineq

n.4.

100)

:

F′ i

=F

d−

dco

sθ|r

ij|

F1

+d

sinθ

|r⊥|

( rij·r

jk

rij·r

ijF

2+

F3

)

F′ j

=d

cosθ

|rij|

F1

−d

sinθ

|r⊥|

( F2+

rij·r

jk

rij·r

ijF

2+

F3

)

F′ k

=d

sinθ

|r⊥|

F2

whe

reF

1=

Fd−

rij·F

d

rij·r

ijr

ij,

F2

=F

1−

r⊥·F

d

r⊥·r

⊥r⊥

and

F3

=r

ij·F

d

rij·r

ijr⊥ (4.1

01)

3out

.A

sa

non-

linea

rco

mbi

natio

nof

thre

eat

oms,

outo

fpla

ne(F

ig.

4.16

3out

):

rd

=r

i+ar

ij+br

ik+c(

rij×

rik

)(4

.102

)

this

enab

les

the

cons

truc

tion

ofdu

mm

ies

outo

fthe

plan

eof

the

othe

rat

oms.

The

forc

eon

part

icle

si,j

andk

due

toth

efo

rce

onth

edu

mm

yca

nbe

com

pute

das

:

F′ j

=

a

−cz i

kcy i

k

cz i

ka

−cx

ik

−cy i

kcx

ika

F d

F′ k

=

b

cz i

j−cy i

j

−cz i

jb

cx

ij

cy i

j−cx

ijb

F dF′ i

=F

d−

F′ j−

F′ k

(4.1

03)

4fd.

Fro

mfo

urat

oms,

with

afix

eddi

stan

ce(F

ig.

4.16

4fd)

:

rd

=r

i+c

rij

+ar

jk+br

jl

|rij

+ar

jk+br

jl|

(4.1

04)

inth

isca

seth

edu

mm

yis

ata

dist

ance

of|c|f

romi.

The

forc

eon

part

icle

si,j,k

andl

due

toth

efo

rce

onth

edu

mm

yca

nbe

com

pute

das

:

F′ i

=F

d−γ(F

d−

p)

F′ j

=(1−a−b)γ(F

d−

p)

F′ k

=aγ(F

d−

p)

F′ l

=bγ

(Fd−

p)

whe

re

γ=

c

|rij

+ar

jk+br

jl|

p=

rid·F

d

rid·r

idr

id

(4.1

05)

Page 94: CS USER AL Simulations - GROMACS

74

Ch

ap

ter

4.

Force

field

s

4.6Long

Range

Electrostatics

4.6.1E

wald

summ

ation

The

totalelectrostaticenergy

ofN

particlesand

theperiodic

images

aregiven

by

V=f2 ∑n

x ∑ny ∑n

z ∗

N∑i

N∑j

qi q

j

rij,n

.(4.106)

(nx ,n

y ,nz )

=n

isthe

boxindex

vector,and

thestar

indicatesthat

terms

with

i=j

shouldbe

omitted

when(n

x ,ny ,n

z )=

(0,0,0).

The

distancerij,n

isthe

realdistancebetw

eenthe

chargesand

notthem

inimum

-image.

This

sumis

conditionallyconvergent,butvery

slow.

Ew

aldsum

mation

was

firstintroduced

asa

method

tocalculate

long-rangeinteractions

ofthe

periodicim

agesin

crystals[

54].T

heidea

isto

convertthesingle

slowly

convergingsum

eqn.4.106

intotw

ofastconverging

terms

anda

constantterm:

V=

Vdir

+V

rec +

V0

(4.107)

Vdir

=f2

N∑i,j ∑nx ∑n

y ∑nz ∗qi q

j erfc(βrij,n )

rij,n

(4.108)

Vrec

=f

2πV

N∑i,j

qi q

j ∑mx ∑m

y ∑mz ∗

exp (−(π

m/β

)2+

2πim

·(ri −

rj ) )

m2

(4.109)

V0

=−fβ

√π

N∑i

q2i ,

(4.110)

where

βis

aparam

eterthat

determines

therelative

weight

ofthe

directand

reciprocalsums

andm

=(m

x ,my ,m

z ).In

thisw

ayw

ecan

usea

shortcut-off

(ofthe

orderof

1nm

)in

thedirect

spacesum

anda

shortcut-offinthe

reciprocalspacesum

(e.g.10w

avevectors

ineach

direction).U

nfortunately,the

computationalcostofthe

reciprocalpartofthesum

increasesas

N2

(orN

3/2

with

aslightly

betteralgorithm

)and

itistherefore

notrealisticto

usefor

anylarge

systems.

Using

Ew

ald

Don’tuse

Ew

aldunless

youare

absolutelysure

thisis

whatyou

want-foralm

ostallcasesthe

PM

Em

ethodbelow

willperform

much

better.If

youstillw

antto

employ

classicalEw

aldsum

mation

enterthis

inyour.m

dp

file,iftheside

ofyourbox

isabout3nm

:

cou

lom

btyp

e=

Ew

ald

rvdw

=0

.9rlist

=0

.9rco

ulo

mb

=0

.9fo

urie

rspa

cing

=0

.6e

wa

ld_

rtol

=1

e-5

A.4

.E

nviron

me

ntV

aria

ble

s1

55

A.3.1

Multi-processor

Optim

ization

Ifyou

want

to,you

couldw

riteyour

own

optimized

comm

unication(perhaps

usingspecific

li-braries

foryour

hardware)

insteadof

MP

I.T

hisshould

neverbe

necessaryfor

normal

use(w

ehaven’theard

ofam

oderncom

puterwhere

itisn’tpossibleto

runM

PI),butifyou

absolutelyw

antto

doit,here

aresom

eclues.

The

interfacebetw

eenthe

comm

unicationroutines

andthe

restof

theG

RO

MA

CS

systemis

de-scribed

inthe

file$G

MX

HO

ME

/src/inclu

de

/ne

two

rk.hW

ew

illgivea

shortdescriptionof

thedifferentroutines

below.

externvoid

gmx

tx(intpid,void*buf,intbufsize);

This

routine,w

hencalled

with

thedestination

processornum

ber,a

pointerto

a(byte

ori-ented)transferbuffer,and

thesize

ofthebufferw

illsendthe

buffertothe

indicatedprocessor

(inour

casealw

aysthe

neighboringprocessor).

The

routinedoesnotw

aituntilthetransfer

isfinished.

externvoid

gmx

txw

ait(intpid);T

hisroutine

waits

untiltheprevious,or

theongoing

transmission

isfinished.

externvoid

gmx

txs(intpid,void*buf,intbufsize);

This

routineim

plements

asynchronous

sendby

callingthe

a-synchronousroutine

andthen

thew

ait.Itm

ightcome

inhandy

tocode

thisdifferently.

externvoid

gmx

rx(intpid,void*buf,intbufsize);

externvoid

gmx

rxw

ait(intpid);

externvoid

gmx

rxs(intpid,void*buf,intbufsize);

The

verysam

eroutines

forreceiving

abuffer

andw

aitinguntilthe

receptionis

finished.

externvoid

gmx

init(intpid,intnprocs);T

hisroutine

initializesthe

differentdevicesneeded

todo

thecom

munication.

Ingeneralit

setsup

thecom

munication

hardware

(ifitisaccessible)ordoes

aninitialize

calltothe

lower

levelcomm

unicationsubsystem

.

externvoid

gmx

stat(FILE

*fp,char*m

sg);W

iththis

routinew

ecan

diagnosethe

ongoingcom

munication.

Inthe

currentim

plemen-

tationit

printsthe

variouscontents

ofthe

hardware

comm

unicationregisters

ofthe

(Inteli860)

multiprocessor

boardsto

afile.

A.4

Environm

entVariables

GR

OM

AC

Sprogram

sm

aybe

influencedby

theuse

ofenvironmentvariables.

Firstofall,the

vari-ables

setinthe

GM

XR

Cfile

areessentialfor

runningand

compiling

GR

OM

AC

S.O

thervariables

are:1.D

UM

PNL,dum

pneighbor

list.Ifsetto

apositive

number

thee

ntire

neighborlistis

printedin

thelog

file(m

aybe

many

megabytes).

Mainly

fordebugging

purposes,butm

ayalso

behandy

forporting

toother

platforms.

Page 95: CS USER AL Simulations - GROMACS

15

4A

pp

en

dix

A.

Tech

nic

alD

eta

ils

The

ener

gies

insi

ngle

prec

isio

nar

eac

cura

teup

toth

ela

stde

cim

al,t

hela

ston

eor

two

deci

mal

sof

the

forc

esar

eno

n-si

gnifi

cant

.T

hevi

riali

sle

ssac

cura

teth

anth

efo

rces

,sin

ceth

evi

riali

son

lyon

eor

dero

fmag

nitu

dela

rger

than

the

size

ofea

chel

emen

tin

the

sum

over

alla

tom

s(s

ec.

B.1

).In

mos

tca

ses

this

isno

trea

llya

prob

lem

,si

nce

the

fluct

uatio

nsin

devi

rialc

anbe

2or

ders

ofm

agni

tude

larg

erth

anth

eav

erag

e.In

perio

dic

char

ged

syst

ems

thes

eer

rors

are

ofte

nne

glig

ible

.E

spec

ially

cut-

off’s

for

the

Cou

lom

bin

tera

ctio

nsca

use

larg

eer

rors

inth

een

ergi

es,

forc

esan

dvi

rial.

Eve

nw

hen

usin

ga

reac

tion-

field

orla

ttice

sum

met

hod

the

erro

rsar

ela

rger

than

orco

mpa

rabl

eto

the

erro

rsdu

eto

the

sing

lepr

ecis

ion.

Sin

ceM

Dis

chao

tic,

traj

ecto

ries

with

very

sim

ilar

star

ting

cond

ition

sw

illdi

verg

era

pidl

y,th

edi

verg

ence

isfa

ster

insi

ngle

prec

isio

nth

anin

doub

lepr

ecis

ion.

For

mos

tsi

mul

atio

nssi

ngle

prec

isio

nis

accu

rate

enou

gh.

Inso

me

case

sdo

uble

prec

isio

nis

re-

quire

dto

getr

easo

nabl

ere

sults

:

•no

rmal

mod

ean

alys

is,

for

the

conj

ugat

egr

adie

ntm

inim

izat

ion

and

the

calc

ulat

ion

and

di-

agon

aliz

atio

nof

the

Hes

sian

•ca

lcul

atio

nof

the

cons

trai

ntfo

rce

betw

een

two

larg

egr

oups

ofat

oms

•en

ergy

cons

erva

tion

(thi

sca

non

lybe

done

with

out

tem

pera

ture

coup

ling

and

with

out

cut-

off’s

)

A.3

Por

ting

GR

OM

AC

S

The

GR

OM

AC

Ssy

stem

isde

sign

edw

ithpo

rtab

ility

asa

maj

orde

sign

goal

.H

owev

erth

ere

are

anu

mbe

rof

thin

gsw

eas

sum

eto

bepr

esen

ton

the

syst

emG

RO

MA

CS

isbe

ing

port

edon

.W

eas

sum

eth

efo

llow

ing

feat

ures

:

1.A

UN

IX-li

keop

erat

ing

syst

em(B

SD

4.x

orS

YS

TE

MV

rev.

3or

high

er)

orU

NIX

-like

libra

ries

runn

ing

unde

re.

g.C

ygW

in

2.an

AN

SIC

com

pile

r

3.op

tiona

llya

For

tran

-77

com

pile

ror

For

tran

-90

com

pile

rfo

rfa

ster

(on

som

eco

mpu

ters

)in

ner

loop

rout

ines

4.op

tiona

llyth

eN

asm

asse

mbl

erto

use

the

asse

mbl

yin

nerlo

ops

onx8

6pr

oces

sors

.

The

rear

eso

me

addi

tiona

lfea

ture

sin

the

pack

age

that

requ

ireex

tra

stuf

fto

bepr

esen

t,bu

tit

isch

ecke

dfo

rin

the

confi

gura

tion

scrip

tand

you

will

bew

arne

dif

anyt

hing

impo

rtan

tis

mis

sing

.

Tha

t’sth

ere

quire

men

tsfo

ra

sing

lepr

oces

sor

syst

em.

Ifyo

uw

ant

toco

mpi

leG

RO

MA

CS

for

am

ultip

lepr

oces

sor

envi

ronm

ent

you

also

need

aM

PI

libra

ry(M

essa

ge-P

assi

ngIn

terf

ace)

tope

rfor

mth

epa

ralle

lcom

mun

icat

ion.

Thi

sis

alw

ays

ship

ped

with

supe

rcom

pute

rs,a

ndfo

rw

ork-

stat

ions

you

can

find

links

tofr

eeM

PI

impl

emen

tatio

nsth

roug

hth

eG

RO

MA

CS

hom

epag

eat

ww

w.g

rom

acs.

org.

4.6

.L

on

gR

an

geE

lect

rost

atic

s7

5

The

fou

rie

rsp

aci

ng

para

met

ertim

esth

ebo

xdi

men

sion

sde

term

ines

the

high

est

mag

nitu

deof

wav

eve

ctor

smx,m

y,m

zto

use

inea

chdi

rect

ion.

With

a3

nmcu

bic

box

this

exam

ple

wou

ldus

e11

wav

eve

ctor

s(f

rom−

5to

5)in

each

dire

ctio

n.T

hee

wa

ldrt

ol

para

met

eris

the

rela

tive

stre

ngth

ofth

eel

ectr

osta

ticin

tera

ctio

nat

the

cut-

off.

Dec

reas

ing

this

give

syo

ua

mor

eac

cura

tedi

rect

sum

,but

ale

ssac

cura

tere

cipr

ocal

sum

.

4.6.

2P

ME

Par

ticle

-mes

hE

wal

dis

am

etho

dpr

opos

edby

Tom

Dar

den

[55

,56]

toim

prov

eth

epe

rfor

man

ceof

the

reci

proc

alsu

m.

Inst

ead

ofdi

rect

lysu

mm

ing

wav

eve

ctor

s,th

ech

arge

sar

eas

sign

edto

agr

idus

ing

card

inal

B-s

plin

ein

terp

olat

ion.

Thi

sgr

idis

then

Fou

rier

tran

sfor

med

with

a3D

FF

Tal

gorit

hman

dth

ere

cipr

ocal

ener

gyte

rmob

tain

edby

asi

ngle

sum

over

the

grid

ink-

spac

e.

The

pote

ntia

lat

the

grid

poin

tsis

calc

ulat

edby

inve

rse

tran

sfor

mat

ion,

and

byus

ing

the

inte

rpo-

latio

nfa

ctor

sw

ege

tthe

forc

eson

each

atom

.

The

PM

Eal

gorit

hmsc

ales

asNlo

g(N

),an

dis

subs

tant

ially

fast

erth

anor

dina

ryE

wal

dsu

mm

a-tio

non

med

ium

tola

rge

syst

ems.

On

very

smal

lsys

tem

sit

mig

htst

illbe

bette

rto

use

Ew

ald

toav

oid

the

over

head

inse

tting

upgr

ids

and

tran

sfor

ms.

Usi

ngP

ME

Tous

eP

artic

le-m

esh

Ew

ald

sum

mat

ion

inG

RO

MA

CS

,spe

cify

the

follo

win

glin

esin

your

.md

pfil

e:

cou

lom

bty

pe

=P

ME

rvd

w=

0.9

rlis

t=

0.9

rco

ulo

mb

=0

.9fo

urie

rsp

aci

ng

=0

.12

pm

e_

ord

er

=4

ew

ald

_rt

ol

=1

e-5

Inth

isca

seth

efou

rie

rsp

aci

ng

para

met

erde

term

ines

the

max

imum

spac

ing

fort

heF

FT

grid

and

pm

eo

rde

rco

ntro

lsth

ein

terp

olat

ion

orde

r.U

sing

4th

orde

r(c

ubic

)in

terp

olat

ion

and

this

spac

ing

shou

ldgi

veel

ectr

osta

ticen

ergi

esac

cura

teto

abou

t5·1

0−3.

Sin

ceth

eLe

nnar

d-Jo

nes

ener

gies

are

nott

his

accu

rate

itm

ight

even

bepo

ssib

leto

incr

ease

this

spac

ing

slig

htly

.

Pre

ssur

esc

alin

gw

orks

with

PM

E,b

utbe

awar

eof

the

fact

that

anis

otro

pic

scal

ing

can

intr

oduc

ear

tifici

alor

derin

gin

som

esy

stem

s.

4.6.

3P

PP

M

The

Par

ticle

-Par

ticle

Par

ticle

-Mes

hm

etho

dsof

Hoc

kney

&E

astw

ood

can

also

beap

plie

din

GR

O-

MA

CS

for

the

trea

tmen

tofl

ong

rang

eel

ectr

osta

ticin

tera

ctio

ns[

57,5

5,58

].W

ithth

isal

gorit

hmth

ech

arge

sof

all

part

icle

sar

esp

read

over

agr

idof

dim

ensi

ons

(n

x,n

y,n

z)

usin

ga

wei

ghtin

g

Page 96: CS USER AL Simulations - GROMACS

76

Ch

ap

ter

4.

Force

field

s

functioncalled

thetriangle-shaped

chargeddistribution:

W(r)

=W

(x)W

(y)W

(z)

W(ξ)

= 34− (

ξh )2

|ξ|≤h2

12 (32−

|ξ|h )2

h2<|ξ|

<3h2

03h2≤|ξ|

(4.111)

where

ξ(is

x,y

orz)

isthe

distanceto

agrid

pointin

thecorresponding

dimension.

Only

the27

closestgridpoints

needto

betaken

intoaccountfor

eachcharge.

Then,

thischarge

distributionis

Fourier

transformed

usinga

3Dinverse

FF

Troutine.

InF

ourierspace

aconvolution

with

functionGis

performed:

G(k)

=g(k)ε0 k

2(4.112)

where

gis

theF

ouriertransform

ofthe

chargespread

functiong(r).

This

yieldthe

longrange

potentialφ(k)on

them

esh,w

hichcan

betransform

edusing

aforw

ardF

FT

routineinto

thereal

spacepotential.

Finally

thepotentialand

forcesare

retrievedusing

interpolation[

58].Itis

noteasyto

calculatethe

fulllong-rangevirialtensor

with

PP

PM

,butitispossible

toobtain

thetrace.

This

means

thatthesum

ofthepressure

components

iscorrect(and

thereforethe

isotropicpressure)but

notnecessarilythe

individualpressurecom

ponents!

Using

PP

PM

Touse

theP

PP

Malgorithm

inG

RO

MA

CS

,specifythe

following

linesin

your.m

dp

file:

cou

lom

btyp

e=

PP

PM

rlist=

1.0

rcou

lom

b=

0.8

5rco

ulo

mb

_sw

itch=

0.0

rvdw

=1

.0fo

urie

rspa

cing

=0

.07

5

For

detailson

thesw

itchparam

eterssee

thesection

onm

odifiedlong-range

interactionsin

thism

anual.W

henusing

PP

PM

we

recomm

endto

takeatm

ost0.075nm

pergridpoint(e.g.20

grid-points

for1.5

nm).

PP

PM

doesnot

providethe

same

accuracyas

PM

Ebut

canbe

slightlyfaster

insom

ecases.

Due

tothe

problemw

iththe

pressuretensor

youshouldn’t

useit

with

pressurecoupling.

We’re

somew

hatambivalentaboutP

PP

M,so

ifyouuse

itpleasecontactus

-otherw

iseitm

ightberem

ovedfrom

futurerelases

sow

ecan

concentrateour

effortson

PM

E.

4.6.4O

ptimizing

Fourier

transforms

Togetthe

bestpossibleperform

anceyou

shouldtry

toavoid

largeprim

enum

bersfor

griddim

en-sions.

The

FF

Tcode

usedin

GR

OM

AC

Sis

optimized

forgrid

sizesofthe

form2

a3b5

c7d11

e13f,

Appendix

A

TechnicalDetails

A.1

Installation

The

entireG

RO

MA

CS

packageis

Free

Softw

are,licensedunderthe

GN

UG

eneralPublic

License.T

hem

aindistribution

siteis

ourW

WW

serveratw

ww

.gromacs.org.

The

packageis

mainly

distributedas

sourcecode,

butwe

alsoprovide

RP

Mpackages

forLinux.

On

thehom

epage

youw

illfindallthe

information

youneed

toinstallthe

package,m

ailinglists

with

archives,andseveraladditionalonline

resourceslike

contributedtopologies,etc.

The

defaultinstallation

actionis

simply

tounpack

thesource

codeand

theissue

./con

figu

rem

ake

ma

kein

stall

The

configurationscript

shouldautom

aticallydeterm

inethe

bestoptions

foryour

platform,

andit

will

tellyou

ifanything

ism

issingon

yoursystem

.You

will

alsofind

detailedstep-by-step

installationinstructions

onthe

website.

A.2

Single

orD

oubleprecision

GR

OM

AC

Scan

becom

piledin

eithersingle

ordouble

precision.T

hedefault

choiceis

singleprecision,butitis

easyto

turnon

doubleprecision

byselecting

the--e

na

ble

-do

ub

leoption

tothe

configurationscript.

Double

precisionw

illbe0

to50%

slowerthan

singleprecision

dependingon

thearchitecture

youare

runningon.

Double

precisionw

illusesom

ewhat

more

mem

oryand

runinput,

energyand

full-precisiontrajectory

filesw

illbealm

osttw

iceas

large.N

otethat

theassem

blyloops

areonly

availablein

singleprecision;

Although

theIntel

SS

E2

instructionset

(availableon

Pentium

IVand

later)supports

doubleprecision

instructionsthe

performance

ism

uchlow

erthan

singleprecision.

Itwould

alsom

eanvery

much

extraw

orkfor

afeature

thatveryfew

peopleuse,

sow

ew

illprobably

notprovide

doubleprecision

assembly

loopsin

thefuture

either.

Page 97: CS USER AL Simulations - GROMACS

15

2C

ha

pte

r8

.A

na

lysi

s4

.7.

All-

hyd

roge

nfo

rce

field

77

whe

ree+f

is0

or1

and

the

othe

rex

pone

nts

arbi

trar

y.(S

eefu

rthe

rth

edo

cum

enta

tion

ofth

eF

FT

algo

rithm

sat

ww

w.ff

tw.o

rg.

Itis

also

poss

ible

toop

timiz

eth

etr

ansf

orm

sfo

rth

ecu

rren

tpro

blem

bype

rfor

min

gso

me

calc

ula-

tions

atth

est

arto

fthe

run.

Thi

sis

notd

one

per

defa

ults

ince

itta

kes

aco

uple

ofm

inut

es,b

utfo

rla

rge

runs

itw

illsa

vetim

e.T

urn

iton

bysp

ecify

ing

op

timiz

e_

fft

=ye

s

inyo

ur.m

dp

file.

Whe

nru

nnin

gin

para

llelt

hegr

idm

ust

beco

mm

unic

ated

seve

ralt

imes

and

thus

hurt

ing

scal

ing

perf

orm

ance

.W

ithP

ME

you

can

impr

ove

this

byin

crea

sing

grid

spac

ing

whi

lesi

mul

tane

ousl

yin

crea

sing

the

inte

rpol

atio

nto

e.g.

6th

orde

r.S

ince

the

inte

rpol

atio

nis

entir

ely

loca

lath

isw

illim

prov

eth

esc

alin

gin

mos

tcas

es.

4.7

All-

hydr

ogen

forc

efiel

d

The

GR

OM

AC

Sal

l-hyd

roge

nfo

rcefi

eld

isal

mos

tide

ntic

alto

the

norm

alG

RO

MA

CS

forc

efiel

d,si

nce

the

extr

ahy

drog

ens

have

noLe

nnar

d-Jo

nes

inte

ract

ion

and

zero

char

ge.

The

only

diffe

renc

esar

ein

the

bond

angl

ean

dim

prop

erdi

hedr

alan

gle

term

s.T

his

forc

efiel

dis

only

usef

ulw

hen

you

need

the

exac

thy

drog

enpo

sitio

ns,

for

inst

ance

for

dist

ance

rest

rain

tsde

rived

from

NM

Rm

easu

rem

ents

.

4.8

GR

OM

OS

-96

note

s

4.8.

1T

heG

RO

MO

S-9

6fo

rce

field

GR

OM

AC

Ssu

ppor

tsth

eG

RO

MO

S-9

6fo

rce

field

s[

46].

All

para

met

ers

for

the

43a1

,43

a2(d

e-ve

lopm

ent,

impr

oved

alka

nedi

hedr

als)

and

43b1

(vac

uum

)fo

rce

field

sar

ein

clud

ed.

All

stan

-da

rdbu

ildin

gbl

ocks

are

incl

uded

and

topo

logi

esca

nbe

build

auto

mat

ical

lyby

pd

b2

gm

x.

The

GR

OM

OS

-96

forc

efie

ldis

afu

rthe

rde

velo

pmen

tof

the

GR

OM

OS

-87

forc

efie

ldon

whi

chth

eG

RO

MA

CS

forc

efiel

dis

base

d.T

heG

RO

MO

S-9

6fo

rce

field

has

impr

ovem

ents

over

the

GR

O-

MA

CS

forc

efie

ldfo

rpr

otei

nsan

dsm

allm

olec

ules

.It

is,

how

ever

,no

tre

com

men

ded

tobe

used

for

long

alka

nes

and

lipid

s.T

heG

RO

MO

S-9

6fo

rce

field

diffe

rsfr

omth

eG

RO

MA

CS

forc

efie

ldin

afe

was

pect

s:

•th

efo

rce

field

para

met

ers

•th

epa

ram

eter

sfo

rth

ebo

nded

inte

ract

ions

are

notl

inke

dto

atom

type

s

•a

four

thpo

wer

bond

stre

tchi

ngpo

tent

ial(

sec.

4.2.

1)

•an

angl

epo

tent

ialb

ased

onth

eco

sine

ofth

ean

gle

(sec

.4.

2.4)

The

rear

etw

odi

ffere

nces

inim

plem

enta

tion

betw

een

GR

OM

AC

San

dG

RO

MO

S-9

6w

hich

can

lead

tosl

ight

lydi

ffere

ntre

sults

whe

nsi

mul

atin

gth

esa

me

syst

emw

ithbo

thpa

ckag

es:

Page 98: CS USER AL Simulations - GROMACS

78

Ch

ap

ter

4.

Force

field

s

•in

GR

OM

OS

-96neighborsearching

forsolventsis

performed

onthe

firstatomofthe

solventm

olecule,this

isnot

implem

entedin

GR

OM

AC

S,

butthe

differencew

ithsearching

with

centersofcharge

groupsis

verysm

all

•the

virialinG

RO

MO

S-96

ism

oleculebased,this

isnotim

plemented

inG

RO

MA

CS

,which

usesatom

icvirials

The

GR

OM

OS

-96force

fieldw

asparam

eterizedw

itha

Lennard-Jonescut-offof1.4

nm,so

besure

touse

aLennard-Jones

cut-offof

atleast

1.4.A

largercut-off

ispossible,

becausethe

Lennard-Jones

potentialandforces

arealm

ostzerobeyond

1.4nm

.

4.8.2G

RO

MO

S-96

files

GR

OM

AC

Scan

readand

write

GR

OM

OS

-96coordinate

andtrajectory

files.T

hesefiles

shouldhave

theextension.g

96

.S

ucha

filecan

bea

GR

OM

OS

-96initial/finalconfiguration

fileor

acoordinate

trajectoryfile

ora

combination

ofboth.T

hefile

isfixed

format,allfloats

arew

rittenas

15.9(files

cangethuge).

GR

OM

AC

Ssupports

thefollow

ingdata

blocksin

thegiven

order:

•H

eaderblock:

TIT

LE

(ma

nd

ato

ry)

•F

rame

blocks:

TIM

ES

TE

P(o

ptio

na

l)P

OS

ITIO

N/P

OS

ITIO

NR

ED

(ma

nd

ato

ry)V

EL

OC

ITY

/VE

LO

CIT

YR

ED

(op

tion

al)

BO

X(o

ptio

na

l)

See

theG

RO

MO

S-96

manual[

46]for

acom

pletedescription

ofthe

blocks.N

otethat

allGR

O-

MA

CS

programs

canread

compressed

orgzip:ed

files.

8.1

3.

Ch

em

icalsh

ifts1

51

thedouble

integralofthecharge

density(

ρ(z)):

ψ(z)−

ψ(−∞

)=− ∫

z

−∞dz′ ∫

z′

−∞ρ(z

′′)dz′′/ε0

(8.38)

where

thepositionz

=−∞

isfar

enoughin

thebulk

phasethat

thefield

iszero.

With

thism

ethod,itispossible

to“split”

thetotalpotentialinto

separatecontributions

fromlipid

andw

aterm

olecules.T

heprogramg

po

ten

tial

dividesthe

boxin

slicesand

sums

allcharges

ofthe

atoms

ineach

slice.Itthen

integratesthis

chargedensity,giving

theelectric

field,andthe

electricfield,giving

thepotential.

Charge

density,fieldand

potentialarew

rittento

xvgr-

inputfiles.

The

programg

coo

rdis

avery

simple

analysisprogram

.A

llitdoes

isprint

thecoordinates

ofselected

atoms

tothree

files,containing

respectivelythe

x-,y-and

z-coordinatesof

thoseatom

s.Itcan

alsocalculate

thecenter

ofmass

ofoneor

more

molecules

andprintthe

coordinatesofthe

centerof

mass

tothree

files.B

yitself,

thisis

probablynot

avery

usefulanalysis,but

havingthe

coordinatesof

selectedm

oleculesor

atoms

canbe

veryhandy

forfurther

analysis,not

onlyin

interfacesystem

s.

The

programg

pvd

calculatesa

lotofproperties,among

which

thedensity

ofagroup

inparticles

perunitofvolum

e,butnota

densitythattakes

them

assofthe

atoms

intoaccount.

The

programg

de

nsity

alsocalculates

thedensity

ofagroup,buttakes

them

assesinto

accountandgives

aplot

ofthe

densityagainst

abox

axis.T

hisis

usefulforlooking

atthe

distributionof

groupsor

atoms

acrossthe

interface.

8.13C

hemicalshifts

tota

ld

o_

shift

Youcan

compute

theN

MR

chemicalshifts

ofprotons

with

theprogram

do

shift

.T

hisis

justan

GR

OM

AC

Sinterface

tothe

publicdom

ainprogram

tota

l[73].

For

furtherinform

ation,readthe

article.

Page 99: CS USER AL Simulations - GROMACS

15

0C

ha

pte

r8

.A

na

lysi

s

HP

r-A

HIS

-15+

THR-16

AR

G-17+

PRO-18

ALA-1

9

ALA-20

GLN-21

PH

E-2

2

VAL-23

LYS

-24+

GLU-25-

ALA

-26

LYS-27+

GLY-28

Fig

ure

8.13

:H

elic

alw

heel

proj

ectio

nof

the

N-t

erm

inal

helix

ofH

Pr.

8.12

Inte

rfac

ere

late

dite

ms

g_

ord

er

g_

de

nsi

tyg

_p

ote

ntia

lg

_co

ord

Whe

nsi

mul

atin

gm

olec

ules

with

long

carb

onta

ils,i

tcan

bein

tere

stin

gto

calc

ulat

eth

eir

aver

age

orie

ntat

ion.

The

rear

ese

vera

lflav

ors

ofor

der

para

met

ers,

mos

tofw

hich

are

rela

ted.

The

prog

ram

go

rde

rca

nca

lcul

ate

orde

rpa

ram

eter

sus

ing

the

equa

tion

Sz

=3 2〈c

os2θ z〉−

1 2(8

.37)

whe

reθ z

isth

ean

gle

betw

een

thez-

axis

ofth

esi

mul

atio

nbo

xan

dth

em

olec

ular

axis

unde

rco

nsid

erat

ion.

The

latte

ris

defin

edas

the

vect

orfr

omC

n−

1to

Cn+

1.

The

para

met

ersSx

andS

yar

ede

fined

inth

esa

me

way

.T

hebr

acke

tsim

ply

aver

agin

gov

ertim

ean

dm

olec

ules

.O

rder

para

met

ers

can

vary

betw

een

1(f

ullo

rder

alon

gth

ein

terf

ace

norm

al)

and

−1/

2(f

ullo

rder

perp

endi

cula

rto

the

norm

al),

with

ava

lue

ofze

roin

the

case

ofis

otro

pic

orie

ntat

ion.

The

prog

ram

can

dotw

oth

ings

fory

ou.

Itca

nca

lcul

ate

the

orde

rpar

amet

erfo

reac

hC

H2

segm

ent

sepa

rate

ly,f

oran

yof

thre

eax

es,o

rit

can

divi

deth

ebo

xin

slic

esan

dca

lcul

ate

the

aver

age

valu

eof

the

orde

rpa

ram

eter

per

segm

enti

non

esl

ice.

The

first

met

hod

give

san

idea

ofth

eor

derin

gof

am

olec

ule

from

head

tota

il,th

ese

cond

met

hod

give

san

idea

ofth

eor

derin

gas

func

tion

ofth

ebo

xle

ngth

.

The

elec

tros

tatic

pote

ntia

l(ψ)a

cros

sth

ein

terf

ace

can

beco

mpu

ted

from

atr

ajec

tory

byev

alua

ting

Cha

pter

5

Topo

logi

es

5.1

Intro

duct

ion

GR

OM

AC

Sm

ust

know

onw

hich

atom

san

dco

mbi

natio

nsof

atom

sth

eva

rious

cont

ribut

ions

toth

epo

tent

ial

func

tions

(see

chap

ter

4)m

ust

act.

Itm

ust

also

know

wha

tpa

ram

eter

sm

ust

beap

plie

dto

the

vario

usfu

nctio

ns.

All

this

isde

scrib

edin

the

top

olo

gyfi

le*.

top

,w

hich

lists

the

con

sta

nt

attribu

teso

fea

chat

om.

The

rear

em

any

mor

eat

omty

pes

than

elem

ents

,bu

ton

lyat

omty

pes

pres

enti

nbi

olog

ical

syst

ems

are

para

met

eriz

edin

the

forc

efie

ld,p

lus

som

em

etal

s,io

nsan

dsi

licon

.T

hebo

nded

and

spec

iali

nter

actio

nsar

ede

term

ined

byfix

edlis

tsth

atar

ein

clud

edin

the

topo

logy

file.

Cer

tain

non-

bond

edin

tera

ctio

nsm

ust

beex

clud

ed(fi

rst

and

seco

ndne

ighb

ors)

,as

thes

ear

eal

read

ytr

eate

din

bond

edin

tera

ctio

ns.

Inad

ditio

nth

ere

are

dyn

am

ica

ttribu

teso

fato

ms:

thei

rpo

sitio

ns,v

eloc

ities

and

forc

es,b

utth

ese

dono

tstr

ictly

belo

ngto

the

mol

ecul

arto

polo

gy.

Thi

sC

hapt

erde

scrib

esth

ese

tup

ofth

eto

polo

gyfil

e,th

e*.

top

file:

wha

tthe

para

met

ers

stan

dfo

ran

dho

w/w

here

toch

ange

them

ifne

eded

.

Not

e:if

you

cons

truc

tyo

urow

nto

polo

gies

,w

een

cour

age

you

toup

load

them

toou

rto

polo

gyar

chiv

eat

ww

w.g

rom

acs.

org!

Just

imag

ine

how

than

kful

you’

dha

vebe

enif

your

topo

logy

had

been

avai

labl

eth

ere

befo

reyo

ust

arte

d.T

hesa

me

goes

for

new

forc

efie

ldor

mod

ified

vers

ions

ofth

est

anda

rdfo

rce

field

s-

cont

ribut

eth

emto

the

forc

efie

ldar

chiv

e!

The

files

are

grou

ped

per

forc

efie

ldty

pe(n

amed

e.g.

gm

xfo

rth

eG

RO

MA

CS

forc

efie

ldor

G4

3a

1fo

rth

eG

RO

MO

S96

forc

efie

ld).

All

files

for

one

forc

efie

ldha

vena

mes

begi

nnin

gw

ithff?

??

whe

re?

??

stan

dsfo

rth

efo

rce

field

nam

e.

5.2

Par

ticle

type

InG

RO

MA

CS

ther

ear

e5

type

sof

part

icle

s,se

eTa

ble

5.1.

Onl

yre

gula

rat

oms

and

dum

my

part

icle

sar

eus

edin

GR

OM

AC

S,n

ucle

i,sh

ells

and

bond

shel

lsar

ene

cess

ary

for

pola

rizab

lefo

rce

field

s,w

hich

we

don’

tyet

have

.

Page 100: CS USER AL Simulations - GROMACS

80

Ch

ap

ter

5.

Top

olog

ies

Particle

Sym

bolatom

sA

nucleussN

shellsS

bondshells

Bdum

mys

D

Table5.1:

Particle

typesin

GR

OM

AC

S

5.2.1A

tomtypes

GR

OM

AC

Suses

47different

atomtypes,

aslisted

below,

with

theircorresponding

masses

(ina.m

.u.).T

hisis

thesam

elisting

asin

thefileff?

??

.atp

(.atp=

atomtype

parameter

file),therefore

inthis

fileyou

canchange

and/oradd

anatom

type.

O1

5.9

99

40

;ca

rbo

nyl

oxyg

en

(C=

O)

OM

15

.99

94

0;

carb

oxyl

oxyg

en

(CO

-)O

A1

5.9

99

40

;h

ydro

xylo

xyge

n(O

H)

OW

15

.99

94

0;

wa

ter

oxyg

en

N1

4.0

06

70

;p

ep

tide

nitro

ge

n(N

or

NH

)N

T1

4.0

06

70

;te

rmin

al

nitro

ge

n(N

H2

)N

L1

4.0

06

70

;te

rmin

al

nitro

ge

n(N

H3

)N

R5

14

.00

67

0;

aro

ma

ticN

(5-rin

g,2

bo

nd

s)N

R5

*1

4.0

06

70

;a

rom

atic

N(5

-ring

,3b

on

ds)

NP

14

.00

67

0;

po

rphyrin

nitro

ge

nC

12

.01

10

0;

ba

reca

rbo

n(p

ep

tide

,C=

O,C

-N)

CH

11

3.0

19

00

;a

liph

atic

CH

-gro

up

CH

21

4.0

27

00

;a

liph

atic

CH

2-g

rou

pC

H3

15

.03

50

0;

alip

hatic

CH

3-g

rou

pC

R5

11

3.0

19

00

;a

rom

atic

CH

-gro

up

(5-rin

g),

un

ited

CR

61

13

.01

90

0;

aro

ma

ticC

H-g

rou

p(6

-ring

),u

nite

dC

B1

2.0

11

00

;b

are

carb

on

(5-,6

-ring

)H

1.0

08

00

;h

ydro

ge

nb

on

de

dto

nitro

ge

nH

O1

.00

80

0;

hyd

roxyl

hyd

rog

en

HW

1.0

08

00

;w

ate

rh

ydro

ge

nH

S1

.00

80

0;

hyd

rog

en

bo

nd

ed

tosu

lfur

S3

2.0

60

00

;su

lfur

FE

55

.84

70

0;

iron

ZN

65

.37

00

0;

zinc

NZ

14

.00

67

0;

arg

NH

(NH

2)

NE

14

.00

67

0;

arg

NE

(NH

)P

30

.97

38

0;

ph

osp

ho

rO

S1

5.9

99

40

;su

ga

ro

re

ster

oxyg

en

CS

11

3.0

19

00

;su

ga

rC

H-g

rou

pN

R6

14

.00

67

0;

aro

ma

ticN

(6-rin

g,2

bo

nd

s)N

R6

*1

4.0

06

70

;a

rom

atic

N(6

-ring

,3b

on

ds)

CS

21

4.0

27

00

;su

ga

rC

H2

-gro

up

SI

28

.08

00

0;

silicon

NA

22

.98

98

0;

sod

ium

(1+

)C

L3

5.4

53

00

;ch

lorin

e(1

-)

8.1

1.

Pro

tein

rela

ted

item

s1

49

0100

200300

400500

600700

800900

1000

1 5 10 15

Residue

Tim

e (ps)C

oilB

endT

urnA

-Helix

B-B

ridge

Figure

8.10:A

nalysisofthe

secondarystructure

elements

ofapeptide

intim

e.

C

O

N

CH

R

C

N

H

H

ψφ

Figure

8.11:D

efinitionofthe

dihedralanglesφ

andψ

oftheprotein

backbone.

–180.0

–120.0

–60.00.0

60.0 �

120.0 �

180.0 �

Phi

–180.0

–120.0

–60.0

0.0

60.0

120.0

180.0

Psi

Ram

achandran Plot

Figure

8.12:R

amachandran

plotofasm

allprotein.

Page 101: CS USER AL Simulations - GROMACS

14

8C

ha

pte

r8

.A

na

lysi

s

•T

helif

etim

eof

the

H-b

onds

isca

lcul

ated

from

the

aver

age

over

alla

utoc

orre

latio

nfu

nctio

nsof

the

exis

tenc

efu

nctio

ns(e

ither

0or

1)of

allH

-bon

ds:

C(τ

)=

〈si(t)s i

(t+τ)〉

(8.3

5)

with

s i(t

)={0,1}

for

H-b

ondi

attim

et.

The

inte

gral

ofC

(τ)

give

sa

roug

hes

timat

eof

the

aver

age

H-b

ond

lifet

imeτ H

B: τ H

B=

∫ ∞ 0C

(τ)dτ

(8.3

6)

Bot

hth

ein

tegr

alan

dth

eco

mpl

ete

auto

corr

elat

ion

func

tion

C(τ

)w

illbe

outp

ut,

soth

atm

ore

soph

istic

ated

anal

ysis

(e.g

.usi

ngm

ulti-

expo

nent

ialfi

ts)

can

beus

edto

getb

ette

res

ti-m

ates

forτ

HB

.

•A

nH

-bon

dex

iste

nce

map

can

bege

nera

ted

ofdi

men

sion

s#

H-b

on

ds×

#fr

am

es.

•In

dex

grou

psar

eou

tput

cont

aini

ngth

ean

alyz

edgr

oups

,al

ldo

nor-

hydr

ogen

atom

pairs

and

acce

ptor

atom

sin

thes

egr

oups

,don

or-h

ydro

gen-

acce

ptor

trip

lets

invo

lved

inhy

drog

enbo

nds

betw

een

the

anal

yzed

grou

psan

dal

lsol

vent

atom

sin

volv

edin

inse

rtio

n.

•S

olve

ntin

sert

ion

into

H-b

onds

can

bean

alyz

ed,s

eeF

ig.

8.9.

Inth

isca

sean

addi

tiona

lgro

upid

entif

ying

the

solv

ent

mus

tbe

sele

cted

.T

heoc

curr

ence

ofin

sert

ion

will

bein

dica

ted

inth

eex

iste

nce

map

.N

ote

that

inse

rtio

nin

toan

dex

iste

nce

ofa

spec

ific

H-b

ond

can

occu

rsi

mul

tane

ousl

yan

dw

illal

sobe

indi

cate

das

such

inth

eex

iste

nce

map

.

8.11

Pro

tein

rela

ted

item

s

do

_d

ssp

g_

ram

axr

am

aw

he

el

Toan

alyz

est

ruct

ural

chan

ges

ofa

prot

ein,

you

can

calc

ulat

eth

era

dius

ofgy

ratio

nor

the

min

imum

resi

due

dist

ance

sdu

ring

time

(see

sec.

8.7)

,or

calc

ulat

eth

eR

MS

D(s

ec.8.8)

.

You

can

also

look

atth

ech

angi

ngofse

con

da

ryst

ruct

ure

ele

me

ntsdu

ring

your

run.

For

this

you

can

use

the

prog

ramd

od

ssp

,whi

chis

anin

terf

ace

for

the

com

mer

cial

prog

ramdssp

[72]

.F

orfu

rthe

rinf

orm

atio

n,se

eth

edssp

-man

ual.

Aty

pica

lout

putp

loto

fdod

ssp

isgi

ven

inF

ig.8

.10.

One

othe

rim

port

anta

naly

sis

ofpr

otei

nsis

the

soca

lled

Ra

ma

cha

nd

ran

plo

t.T

his

isth

epr

ojec

tion

ofth

est

ruct

ure

onth

etw

odi

hedr

alan

glesφ

andψ

ofth

epr

otei

nba

ckbo

ne,s

eeF

ig.

8.11

.

Toev

alua

teth

isR

amac

hand

ran

plot

you

can

use

the

prog

ram

gra

ma

.A

typi

calo

utpu

tis

give

nin

Fig

.8.1

2.

Itis

also

poss

ible

toge

nera

tean

anim

atio

nof

the

Ram

acha

ndra

npl

otin

time.

Thi

sca

nbe

ofhe

lpfo

ran

alyz

ing

cert

ain

dihe

dral

tran

sitio

nsin

your

prot

ein.

You

can

use

the

prog

ram

xra

ma

for

this

.

Whe

nst

udyi

ngα

-hel

ices

itis

usef

ulto

have

ahelic

al

wh

ee

lpro

ject

ion

ofyo

urpe

ptid

e,to

see

whe

ther

ape

ptid

eis

amph

ipat

ic.

Thi

sca

nbe

done

usin

gth

ew

he

el

prog

ram

.Tw

oex

ampl

esar

epl

otte

din

Fig

.8.1

3.

5.2

.P

art

icle

typ

e8

1

CA

40

.08

00

0;

calc

ium

(2+

)M

G2

4.3

05

00

;m

ag

ne

sium

(2+

)F

18

.99

84

0;

fluo

rine

(co

v.b

ou

nd

)C

P2

14

.02

70

0;

alip

ha

ticC

H2

-gro

up

usi

ng

Ryc

kae

rt-B

ell.

CP

31

5.0

35

00

;a

liph

atic

CH

3-g

rou

pu

sin

gR

ycka

ert

-Be

ll.C

R5

12

.01

10

0;

aro

ma

ticC

H-g

rou

p(5

-rin

g)+

HC

R6

12

.01

10

0;

aro

ma

ticC

-b

on

de

dto

H(6

-rin

g)+

HH

CR

1.0

08

00

;H

atta

che

dto

aro

ma

ticC

(5o

r6

riO

WT

31

5.9

99

40

;T

IP3

Pw

ate

ro

xyg

en

SD

32

.06

00

0;

DM

SO

Su

lph

ur

OD

15

.99

94

0;

DM

SO

Oxy

ge

nC

D1

5.0

35

00

;D

MS

OC

arb

on

Ato

mic

deta

ilis

used

exce

ptfo

rhy

drog

enat

oms

boun

dto

(alip

hatic

)ca

rbon

atom

s,w

hich

are

trea

ted

asu

nite

da

tom

s.N

osp

ecia

lhyd

roge

n-bo

ndte

rmis

incl

uded

.

The

last

10at

omty

pes

are

extr

aat

omty

pes

with

resp

ectt

oth

eG

RO

MO

S-8

7fo

rce

field

[39

]:

•F

was

take

nfr

omre

f.[43

],

•C

P2

and

CP

3fr

omre

f.[40]a

ndre

fere

nces

cite

dth

erei

n,

•C

R5,

CR

6an

dH

CR

from

ref.

[59]

•O

WT

3fr

omre

f.[4

2]

•S

D,O

Dan

dC

Dfr

omre

f.[44

]

The

refo

re,i

fyou

use

the

GR

OM

AC

Sfo

rce

field

asit

is,m

ake

sure

you

use

the

refe

renc

esin

your

publ

icat

ions

asm

entio

ned

abov

e.

Not

e:G

RO

MA

CS

mak

esus

eof

the

atom

type

sas

ana

me,

no

tas

anu

mbe

r(a

se.

g.in

GR

OM

OS

).

5.2.

2D

umm

yat

oms

Som

efo

rce

field

sus

edu

mm

yat

oms

(virt

uals

ites

that

are

cons

truc

ted

from

real

atom

s)on

whi

chce

rtai

nin

tera

ctio

nsar

elo

cate

d(e

.g.o

nbe

nzen

erin

gs,t

ore

prod

uce

the

corr

ectq

uadr

upol

e).

Thi

sis

desc

ribed

inse

c.4.5

.

Tom

ake

dum

my

atom

sin

your

syst

em,

you

shou

ldin

clud

ea

sect

ion

[d

um

mie

s?]

inyo

urto

polo

gyfil

e,w

here

the

‘?’

stan

dsfo

rth

enu

mbe

rco

nstr

uctin

gat

oms

for

the

dum

my

atom

.T

his

will

be‘2

’for

type

2,‘3

’for

type

s3,

3fd,

3fad

and

3out

and

‘4

’for

type

4fd

(the

diffe

rent

type

sar

eex

plai

ned

inse

c.4.5

).

Par

amet

ers

for

type

2sh

ould

look

like

this

:

[d

um

mie

s2]

;D

um

my

fro

mfu

nct

a5

12

10

.74

39

75

6

for

type

3lik

eth

is:

Page 102: CS USER AL Simulations - GROMACS

82

Ch

ap

ter

5.

Top

olog

ies

[d

um

mie

s3]

;D

um

my

from

fun

cta

b5

12

31

0.7

43

97

56

0.1

28

01

2

fortype

3fdlike

this:

[d

um

mie

s3]

;D

um

my

from

fun

cta

d5

12

32

0.5

-0.1

05

fortype

3fadlike

this:

[d

um

mie

s3]

;D

um

my

from

fun

ctth

eta

d5

12

33

12

00

.5

fortype

3outlikethis:

[d

um

mie

s3]

;D

um

my

from

fun

cta

bc

51

23

4-0

.4-0

.46

.92

81

fortype

4fdlike

this:

[d

um

mie

s4]

;D

um

my

from

fun

cta

bd

51

23

41

0.3

33

33

0.3

33

33

-0.1

05

This

willresultin

theconstruction

ofadum

my

‘atom’,num

ber5(firstcolum

n‘

Du

mm

y’),basedon

thepositions

of1and

2or

1,2and

3or

1,2,3and

4(nexttw

o,threeor

fourcolum

ns‘

from

’)fol-

lowing

therules

determined

bythe

functionnum

ber(nextcolum

n‘

fun

ct’)

with

theparam

etersspecified

(lastone,two

orthree

columns

‘a

b.

.’).

Note

thatany

bondsdefined

between

dumm

yatom

sand/or

normal

atoms

will

berem

ovedby

gro

mp

pafter

theexclusions

havebeen

generated.T

hisw

ay,exclusionsw

illnotbeaffected

byan

atombeing

definedas

dumm

yatom

ornot,butby

thebonding

configurationofthe

atom.

5.3P

arameter

files

5.3.1A

toms

Anum

berofsta

ticproperties

areassigned

tothe

atomtypes

inthe

GR

OM

AC

Sforce

field:Type,

Mass,

Charge,εand

σ(see

Table5.2T

hem

assis

listedinff?

??

.atp

(see5.2.1),w

hereasthe

chargeis

listedinff?

??

.rtp(.rtp

=residuetopologyparam

eterfile,

see5.5.1).T

hisim

pliesthat

thecharges

areonly

definedin

thebuilding

blocksof

amino

acidsor

userdefined

buildingblocks.

When

generatinga

topology(

*.top

)using

thepd

b2

gm

xprogram

theinform

ationfrom

thesefiles

iscom

bined.

The

following

dyn

am

icquantitiesare

associatedw

ithan

atom

8.1

0.

Hyd

rogen

bo

nd

s1

47

D

H

α

A

r

Figure

8.8:G

eometricalH

ydrogenbond

criterion.

O

DA

H

H

H

(1)(2)

(2)

Figure

8.9:Insertion

ofw

aterinto

anH

-bond.(1)

Norm

alH

-bondbetw

eentw

oresidues.

(2)H

-bondingbridge

viaa

water

molecule.

8.10H

ydrogenbonds

g_

hb

on

d

The

programg

hb

on

danalyses

thehyd

rogen

bo

nd

s(H-bonds)betw

eenallpossible

donorsD

andacceptors

A.To

determine

ifanH

-bondexists,a

geometricalcriterion

isused,see

alsoF

ig.8.8:

r≤

rH

B=

0.35nm

α≤

αH

B=

60o

(8.34)

The

valueofr

HB

=0.35

nmcorresponds

tothe

firstminim

umofthe

rdfofSP

C-w

ater(see

alsoF

ig.8.3).

The

programg

hb

on

danalyses

allhydrogenbonds

existingbetw

eentw

ogroups

ofatoms

(which

mustbe

eitheridenticalor

non-overlapping)or

inspecified

Donor

Hydrogen

Acceptor

triplets,inthe

following

ways:

•D

onor-Acceptor

distance(r)

distributionofallH

-bonds

•H

ydrogen-Donor-A

cceptorangle

)distribution

ofallH-bonds

•T

hetotalnum

berofH

-bondsin

eachtim

efram

e

•T

henum

berof

H-bonds

intim

ebetw

eenresidues,

dividedinto

groupsn

-n+i

where

nand

n+i

standfor

residuenum

bersandigoes

from0

to6.

The

groupfori=

6also

includesallH

-bondsfori

>6.

These

groupsinclude

then-n+3,

n-n

+4and

n-n

+5H

-bondsw

hichprovide

am

easurefor

theform

ationofα

-helicesorβ

-turnsor

strands.

Page 103: CS USER AL Simulations - GROMACS

14

6C

ha

pte

r8

.A

na

lysi

s

beca

lcul

ated

from

the

eige

nval

ues

λi

and

the

eige

nvec

tors

,whi

char

eth

eco

lum

nsof

the

rota

tion

mat

rixR

.F

ora

sym

met

rican

ddi

agon

ally

-dom

inan

tm

atrixA

ofsi

ze3N

×3N

the

squa

rero

otca

nbe

calc

ulat

edas

:

A1 2

=R

diag

(λ1 2 1,λ

1 2 2,...,λ

1 2 3N

)RT

(8.2

7)

Itca

nbe

verifi

edea

sily

that

the

prod

uct

ofth

ism

atrix

with

itsel

fgi

ves

A.

Now

we

can

defin

ea

diffe

renc

edbe

twee

nco

varia

nce

mat

ricesA

andB

asfo

llow

s:

d(A,B

)=

√ tr( ( A

1 2−B

1 2

) 2)(8

.28)

=√ tr

( A+B−

2A1 2B

1 2

)(8

.29)

=

N ∑ i=1

( λA i

B i

) −2

N ∑ i=1

N ∑ j=1

√ λA iλ

B j

( RA i·R

B j

) 2 1 2

(8.3

0)

whe

retr

isth

etr

ace

ofa

mat

rix.

We

can

now

defin

eth

eov

erla

ps

as:

s(A,B

)=

1−

d(A,B

)√

trA

+trB

(8.3

1)

The

over

lap

is1

ifan

don

lyif

mat

ricesA

andB

are

iden

tical

.It

is0

whe

nth

esa

mpl

edsu

bspa

ces

are

com

plet

ely

orth

ogon

al.

Aco

mm

only

used

mea

sure

isth

esu

bspa

ceov

erla

pof

the

first

few

eige

nvec

tors

ofco

varia

nce

mat

rices

.T

heov

erla

pof

the

subs

pace

span

ned

bym

orth

onor

mal

vect

orsw

1,...,w

mw

itha

refe

renc

esu

bspa

cesp

anne

dbynor

thon

orm

alve

ctor

sv1,...,v

nca

nbe

quan

tified

asfo

llow

s:

over

lap(v,w

)=

1 n

n ∑ i=1

m ∑ j=1

(vi·w

j)2

(8.3

2)

The

over

lap

will

incr

ease

with

incr

easi

ngman

dw

illbe

1w

hen

setv

isa

subs

pace

ofse

tw.

The

disa

dvan

tage

ofth

ism

etho

dis

that

itdo

esno

ttak

eth

eei

genv

alue

sin

toac

coun

t.A

llei

genv

ecto

rsar

ew

eigh

ted

equa

llyan

dw

hen

dege

nera

tesu

bspa

ces

are

pres

ent(

equa

leig

enva

lues

)the

calc

ulat

edov

erla

pw

illbe

too

low

.

Ano

ther

usef

ulch

eck

isth

eco

sine

cont

ent.

Itha

sbe

enpr

oven

the

the

prin

cipa

lcom

pone

nts

ofra

ndom

diffu

sion

are

cosi

nes

with

the

num

ber

ofpe

riods

equa

lto

half

the

prin

cipa

lcom

pone

ntin

dex[

71].

The

eige

nval

ues

are

prop

ortio

nalt

oth

ein

dex

toth

epo

wer

−2.

The

cosi

neco

nten

tis

defin

edas

:2 T

( ∫ T 0co

s(kπt)p

i(t)

dt

) 2(∫ T 0

p2 i(t

)dt) −1

(8.3

3)

Whe

nth

eco

sine

cont

ento

fthe

first

few

prin

cipa

lcom

pone

nts

iscl

ose

to1,

the

larg

estfl

uctu

atio

nsar

eno

tcon

nect

edw

ithth

epo

tent

ial,

butw

ithra

ndom

diffu

sion

.

The

cova

rianc

em

atrix

isbu

iltan

ddi

agon

aliz

edbyg

cova

r.

The

prin

cipa

lco

mpo

nent

san

dov

erla

p(a

nym

any

mor

eth

ings

)ca

nbe

plot

ted

and

anal

yzed

with

ga

na

eig

.T

heco

sine

cont

ent

can

beca

lcul

ated

withg

an

aly

ze.

5.3

.P

ara

me

ter

file

s8

3

Pro

pert

yS

ymbo

lU

nit

Type

--

Mas

sm

a.m

.u.

Cha

rge

qel

ectr

onep

silo

kJ/m

olsi

gma

σnm

Tabl

e5.

2:S

tatic

atom

type

prop

ertie

sin

GR

OM

AC

S

•P

ositi

onx

•Ve

loci

tyv

The

sequ

antit

ies

are

liste

din

the

coor

dina

tefil

e,*.

gro

(see

sect

ion

File

form

at,5.6.

6).

5.3.

2B

onde

dpa

ram

eter

s

The

bond

edpa

ram

eter

s(i.

e.bo

nds,

bond

angl

es,

impr

oper

and

prop

erdi

hedr

als)

are

liste

din

ff?

??

bo

n.it

p.

The

term

fun

cis

1fo

rha

rmon

ican

d2

for

GR

OM

OS

-96

bond

and

angl

epo

tent

ials

.F

orth

edi

hedr

al,t

his

isex

plai

ned

afte

rth

islis

ting.

[b

on

dty

pe

s]

;i

jfu

nc

b0

kbC

O1

0.1

23

00

50

20

80

.C

OM

10

.12

50

04

18

40

0.

......

[a

ng

lety

pe

s]

;i

jk

fun

cth

0ct

hH

OO

AC

11

09

.50

03

97

.48

0H

OO

AC

H1

11

09

.50

03

97

.48

0......

[d

ihe

dra

ltyp

es

];

il

fun

cq

0cq

NR

5*

NR

52

0.0

00

16

7.3

60

NR

5*

NR

5*

20

.00

01

67

.36

0......

[d

ihe

dra

ltyp

es

];

jk

fun

cp

hi0

cpm

ult

CO

A1

18

0.0

00

16

.73

62

CN

11

80

.00

03

3.4

72

2......

[d

ihe

dra

ltyp

es

]; ;

Ryc

kae

rt-B

elle

ma

ns

Dih

ed

rals

;

Page 104: CS USER AL Simulations - GROMACS

84

Ch

ap

ter

5.

Top

olog

ies

;a

ja

kfu

nct

CP

2C

P2

39.2

78

91

2.1

56

-13

.12

0-3

.05

97

26

.24

0-3

1.4

95

Also

inthis

fileare

theR

yckaert-Bellem

ans[

60]parameters

forthe

CP

2-CP

2dihedrals

inalkanes

oralkane

tailsw

iththe

following

constants:

(kJ/mol)

C0

=9.28

C2

=−

13.12C

4=

26.24C

1=

12.16C

3=

−3.06

C5

=−

31.5

(Note:T

heuse

ofthispotentialim

pliesexclusions

ofLJ-interactionsbetw

eenthe

firstandthe

lastatom

ofthedihedral,andψ

isdefined

accordingto

the’polym

erconvention’(

ψtr

ans

=0)).

So

thereare

threetypes

ofdihedralsin

theG

RO

MA

CS

forcefield:

•proper

dihedral:funct=

1,with

mult=

multiplicity,so

thenum

berofpossible

angles

•im

properdihedral:

funct=2

•R

yckaert-Bellem

ansdihedral:

funct=3

Inthe

fileff?

??

bo

n.itp

youcan

addbonded

parameters.

Ifyou

want

toinclude

parameters

fornew

atomtypes,m

akesure

youdefine

thisnew

atomtype

inff?

??

.atp

asw

ell.

5.3.3N

on-bondedparam

eters

The

non-bondedparam

etersconsistofthe

Vander

Waals

parameters

Aand

C,as

listedin

thefile

ff??

?n

b.itp

,wherep

type

isthe

particletype

(seeTable5.1):

[a

tom

type

s]

;na

me

ma

ssch

arg

ep

type

c6c1

2O

15

.99

94

00

.00

0A

0.2

26

17

E-0

20

.74

15

8E

-06

OM

15

.99

94

00

.00

0A

0.2

26

17

E-0

20

.74

15

8E

-06

.....

[n

on

bo

nd

_p

ara

ms

];

ij

fun

cc6

c12

OO

10

.22

61

7E

-02

0.7

41

58

E-0

6O

OA

10

.22

61

7E

-02

0.1

38

07

E-0

5.....

[p

airtyp

es

];

ij

fun

ccs6

cs12

;T

HE

SE

AR

E1

-4IN

TE

RA

CT

ION

SO

O1

0.2

26

17

E-0

20.7

41

58

E-0

6O

OM

10

.22

61

7E

-02

0.7

41

58

E-0

6.....

With

Aand

Cbeing

definedas

Aii =

4εi σ

12

i(5.1)

8.9

.C

ovaria

nce

an

alysis

14

5

Instead

ofcomparing

thestructures

tothe

initialstructureattim

et=

0(so

forexam

plea

crystalstructure),one

canalso

calculateeqn.8.20

usinga

time

shiftτ:

RMSD

(t;τ)= [

1N

N∑i=1 ‖r

i (t)−ri (t−

τ)‖2 ]

12

(8.22)

socom

paringto

aleast-square

structureat

t−τ.

This

givessom

einsight

inthe

mobility

asa

functionofτ.

Use

theprogramg

run

rms

.

8.9C

ovarianceanalysis

Covariance

analysis,alsocalled

principalcomponentanalysis

oressentialdynam

ics[

70],canfind

correlatedm

otions.Ituses

thecovariance

matrixC

oftheatom

iccoordinates:

Cij

= ⟨M

12ii (xi −

〈xi 〉)M

12jj (xj −

〈xj 〉) ⟩

(8.23)

where

Mis

adiagonal

matrix

containingthe

masses

ofthe

atoms

(mass-w

eightedanalysis)

orthe

unitm

atrix(non-m

assw

eightedanalysis).

Cis

asym

metric3

3Nm

atrix,w

hichcan

bediagonalized

with

anorthonorm

altransformation

matrixR

:

RTCR

=diag(λ

1 ,λ2 ,...,λ

3N

)w

hereλ

1≥λ

2≥...≥

λ3N

(8.24)

The

columns

ofRare

theeigenvectors,

alsocalled

principalor

essentialm

odes.R

definesa

transformation

toa

newcoordinate

system.

The

trajectorycan

beprojected

onthe

principalmodes

togive

theprincipalcom

ponentspi (t):

p(t)

=R

TM

12(x(t)−〈x〉)

(8.25)

The

eigenvalueλi is

them

eansquare

fluctuationofprincipalcom

ponenti.

The

firstfewprincipal

modes

oftendescribe

collective,globalmotions

inthe

system.

The

trajectorycan

befiltered

alongone

(orm

ore)principalm

odes.F

orone

principalmodei

thisgoes

asfollow

s:

xf(t)

=〈x〉

+M

−12R

∗i p

i (t)(8.26)

When

theanalysis

isperform

edon

am

acromolecule,one

oftenw

antsto

remove

theoverallrota-

tionand

translationto

lookatthe

internalmotion

only.T

hiscan

beachieved

byleastsquare

fittingto

areference

structure.C

arehas

tobe

takenthat

thereference

structureis

representativefor

theensem

ble,since

thechoice

ofreference

structureinfluences

thecovariance

matrix.

One

shouldalw

ayscheck

ifthe

principalmodes

arew

elldefined.If

thefirst

principalcomponent

resembles

ahalf

cosineand

thesecond

resembles

afullcosine,

youm

ightbe

filteringnoise.

Agood

way

tocheck

therelevance

ofthefirstfew

principalmodes

isto

calculatethe

overlapofthe

sampling

between

thefirstand

secondhalfofthe

simulation.

Note

thatthiscan

onlybe

donew

henthe

same

referencestructure

isused

forthe

two

halves.

The

elements

ofthe

covariancem

atrixare

proportionaltothe

squareof

thedisplacem

ent,so

we

needto

takethe

squarerootofthe

matrix

toexam

inethe

extentofsampling.

The

squarerootcan

Page 105: CS USER AL Simulations - GROMACS

14

4C

ha

pte

r8

.A

na

lysi

s

2130

4050

6070

8090

2130405060708090 t=0 ps

Res

idue

Num

ber

0D

ista

nce

(nm

)1.

2

Fig

ure

8.7:

Am

inim

umdi

stan

cem

atrix

for

ape

ptid

e[

3].

8.8

Roo

tmea

nsq

uare

devi

atio

nsin

stru

ctur

e

g_

rms

g_

rmsd

ist

The

roo

tme

an

squ

are

dev

iatio

n(RMSD

)of

cert

ain

atom

sin

am

olec

ule

with

resp

ectt

oa

refe

r-en

cest

ruct

ure

can

beca

lcul

ated

with

the

prog

ram

grm

sby

leas

t-sq

uare

fittin

gth

est

ruct

ure

toth

ere

fere

nce

stru

ctur

e(

t 2=

0)an

dsu

bseq

uent

lyca

lcul

atin

gth

eRMSD

(eqn

.8.2

0).

RMSD

(t1,t

2)

=

[ 1 N

N ∑ i=1

‖ri(t 1

)−

r i(t

2)‖

2

]1 2

(8.2

0)

whe

rer i

(t)

isth

epo

sitio

nof

atomi

attim

et.

NO

TE

that

fittin

gdo

esno

tha

veto

use

the

sam

eat

oms

asth

eca

lcul

atio

nof

theRMSD

;e.

g.:

apr

otei

nis

usua

llyfit

ted

onth

eba

ckbo

neat

oms

(N,C

α,C

),bu

ttheRMSD

can

beco

mpu

ted

ofth

eba

ckbo

neor

ofth

ew

hole

prot

ein.

Inst

ead

ofco

mpa

ring

the

stru

ctur

esto

the

initi

alst

ruct

ure

attim

et

=0

(so

for

exam

ple

acr

ysta

lst

ruct

ure)

,one

can

also

calc

ulat

eeq

n.8.

20w

itha

stru

ctur

eat

timet 2

=t 1−τ.

Thi

sgi

ves

som

ein

sigh

tin

the

mob

ility

asa

func

tion

ofτ.

Als

oa

mat

rixca

nbe

mad

ew

ithth

eRMSD

asa

func

tion

oft 1

andt 2

,thi

sgi

ves

ani

cegr

aphi

cali

mpr

essi

onof

atr

ajec

tory

.If

ther

ear

etr

ansi

tions

ina

traj

ecto

ry,t

hey

will

clea

rlysh

owup

insu

cha

mat

rix.

Alte

rnat

ivel

yth

eRMSD

can

beco

mpu

ted

usin

ga

fit-f

ree

met

hod

with

the

prog

ram

grm

sdis

t:

RMSD

(t)

=

1 N2

N ∑ i=1

N ∑ j=1

‖rij(t

)−

r ij(0

)‖2

1 2

(8.2

1)

whe

reth

edis

tan

cer i

jbe

twee

nat

oms

attim

etis

com

pare

dw

ithth

edi

stan

cebe

twee

nth

esa

me

atom

sat

time0

.

5.4

.C

on

stra

ints

85

Cii

=4ε

iσ6 i

(5.2

)

and

com

pute

dac

cord

ing

toth

eco

mbi

natio

nru

les

:

Aij

=(A

iiA

jj)1 2

(5.3

)

Cij

=(C

iiC

jj)1 2

(5.4

)

Itis

also

poss

ible

tous

eth

eco

mbi

natio

nru

les

whe

reth

’sar

eav

erag

ed:

σij

=1 2(σ

ii+σ

jj)

(5.5

)

ε ij

=√ε i

iεjj

(5.6

)

5.3.

41-

4in

tera

ctio

ns

The

1-4

inte

ract

ions

for

the

atom

type

sinff?

??

nb

.itp

are

liste

din

the[

pa

irty

pe

s]

sect

ion.

The

GR

OM

AC

San

dG

RO

MO

Sfo

rce

field

slis

tsal

lthe

sein

tera

ctio

nsex

plic

itly,

butt

his

sect

ion

mig

htbe

empt

yfo

rfo

rce

field

slik

eO

PLS

that

calc

ulat

eth

e1-

4in

tera

ctio

nsby

scal

ing.

5.3.

5E

xclu

sion

s

The

excl

usio

nsfo

rbo

nded

part

icle

sar

ege

nera

ted

byg

rom

pp

for

neig

hbor

ing

atom

sup

toa

cert

ain

num

ber

ofbo

nds

away

,as

defin

edin

the

[m

ole

cule

typ

e]

sect

ion

inth

eto

polo

gyfil

e(s

ee5.

6.1)

.P

artic

les

are

cons

ider

edbo

nded

whe

nth

eyar

eco

nnec

ted

bybo

nds

([

bo

nd

s]

type

s1

to5)

orco

nstr

aint

s(

[co

nst

rain

ts]

type

1).

[b

on

ds

]ty

pe5

can

beus

edto

crea

tea

conn

ectio

nbe

twee

ntw

oat

oms

with

out

crea

ting

anin

tera

ctio

n.T

here

isa

harm

onic

inte

ract

ion

([b

on

ds

]ty

pe6)

whi

chdo

esno

tco

nnec

tth

eat

oms

bya

chem

ical

bond

.T

here

isal

soa

seco

ndco

nstr

aint

type

([

con

stra

ints

]ty

pe2)

whi

chfix

esth

edi

stan

ce,b

utdo

esno

tcon

nect

the

atom

sby

ach

emic

albo

nd.

Ext

raex

clus

ions

with

ina

mol

ecul

eca

nbe

adde

dm

anua

llyin

a[

exc

lusi

on

s]

sect

ion.

Eac

hlin

esh

ould

star

tw

ithon

eat

omin

dex,

follo

wed

byon

eor

mor

eat

omin

dice

s.A

llno

n-bo

nded

inte

ract

ions

betw

een

the

first

atom

and

the

othe

rat

oms

will

beex

clud

ed.

Whe

nal

lnon

-bon

ded

inte

ract

ions

with

inor

betw

een

grou

psof

atom

sne

edto

beex

clud

ed,

isit

mor

eco

nven

ient

and

muc

hm

ore

effic

ient

tous

een

ergy

mon

itor

grou

pex

clus

ions

(see

sec.

3.3)

.

5.4

Con

stra

ints

Con

stra

ints

are

defin

edin

the[co

nst

rain

ts]

sect

ion.

The

form

atis

two

atom

num

bers

fol-

low

edby

the

func

tion

type

,w

hich

can

be1

or2

and

the

cons

trai

ntdi

stan

ce.

The

only

diffe

renc

ebe

twee

nth

etw

oty

pes

isth

atty

pe1

isus

edfo

rge

nera

ting

excl

usio

nsan

dw

hile

type

2is

not(

see

5.3.

5).

The

dist

ance

sar

eco

nstr

aine

dus

ing

the

LIN

CS

orth

eS

HA

KE

algo

rithm

,whi

chca

nbe

se-

lect

edin

the*

.md

pfil

e.B

oth

type

sof

cons

trai

nts

can

bepe

rtur

bed

infr

ee-e

nerg

yca

lcul

atio

nsby

addi

nga

seco

ndco

nstr

aint

dist

ance

(see

5.6.

5).

Sev

eral

type

sof

bond

san

dan

gles

(see

Tabl

e5.

4)

Page 106: CS USER AL Simulations - GROMACS

86

Ch

ap

ter

5.

Top

olog

ies

canbe

convertedautom

aticallyto

constraintsby

gro

mp

p.

There

areseveraloptions

forthis

inthe

*.md

pfile.

We

havealso

implem

entedthe

SE

TT

LEalgorithm

[27]w

hichis

ananalyticalsolution

ofSH

AK

Especifically

forw

ater.S

ET

TLE

canbe

selectedin

thetopology

file.C

heckfor

instancethe

SP

Cm

oleculedefinition:

[m

ole

cule

type

];

mo

lna

me

nre

xclS

OL

1

[a

tom

s]

;n

ra

ttyp

ere

sn

rre

nn

ma

tn

mcg

nr

cha

rge

1O

W1

SO

LO

W1

1-0

.82

2H

W1

SO

LH

W2

10

.41

3H

W1

SO

LH

W3

10

.41

[se

ttles

];

OW

fun

ctd

oh

dh

h1

10

.10

.16

33

3

[e

xclusio

ns

]1

23

21

33

12

The

section[se

ttles

]defines

thefirstatom

ofthew

aterym

olecule,thesettle

functisalw

aysone,and

thedistance

between

Oand

H,and

distancebetw

eenboth

Hatom

sm

ustbegiven.

Note

thatthe

algorithmcan

alsobe

usedfor

TIP

3Pand

TIP

4P[

42].T

IP3P

justhas

anothergeom

etry.T

IP4P

hasa

dumm

yatom

,butsincethatis

generateditdoes

notneedto

beshaken

(norstirred).

5.5D

atabases

5.5.1R

esiduedatabase

The

fileholding

theresidue

databaseis

ff??

?.rtp

.O

riginallythis

filecontained

buildingblocks

(amino

acids)forproteins,andis

theG

RO

MA

CS

interpretationofthert3

7c4

.da

tfile

ofGR

O-

MO

S.S

othe

residuefile

containsinform

ation(bonds,

charge,charge

groupsand

improper

dihe-drals)

fora

frequentlyused

buildingblock.

Itisbetter

no

ttochange

thisfile

becauseitis

standardinput

forpd

b2

gm

x,

butif

changesare

neededm

akethem

inthe

*.top

file(see

sectionTopol-

ogyfile,5.6.1).

How

ever,in

theff??

?.rtp

filethe

usercan

definea

newbuilding

blockor

molecule:

seefor

example

2,2,2-trifluoroethanol(T

FE

)or

n-decane(C

10).B

utw

hendefining

newm

olecules(non-protein)

itis

preferableto

createa

*.itpfile.

This

willbe

discussedin

anextsection

(section5.6.2).

The

fileff?

??

.rtpis

onlyused

bypd

b2

gm

x.

As

mentioned

before,theonly

extrainform

ationthis

programneeds

fromff??

?.rtp

isbonds,

chargesof

atoms,

chargegroups

andim

proper

8.7

.R

ad

ius

ofg

yratio

na

nd

dista

nce

s1

43

For

planesituses

thenorm

alvectorperpendicular

tothe

plane.Itcan

alsocalculate

thed

istan

cedbetw

eenthe

geometricalcenter

oftwo

planes(see

Fig.

8.6D),and

thedistancesd1

andd

2betw

een2

atoms

(ofa

vector)and

thecenter

ofa

planedefined

by3

atoms

(seeF

ig.8.6D

).It

furthercalculates

thedistancedbetw

eenthe

centerofthe

planeand

them

iddleofthis

vector.D

ependingon

theinput

groups(i.e.

groupsof

2or

3atom

numbers),

theprogram

decidesw

hatangles

anddistances

tocalculate.

For

example,the

index-filecould

looklike

this:

[a

_p

lan

e]

12

3[

a_

vecto

r]

34

5

8.7R

adiusofgyration

anddistances

g_

gyra

teg

_sg

an

gle

g_

min

dist

g_

md

ma

txp

m2

ps

Tohave

arough

measure

forthe

compactness

ofastructure,you

cancalculate

thera

diu

so

fgyra

-tio

nw

iththe

programg

gyra

teas

follows:

Rg

= (∑i ‖r

i ‖2m

i∑

i mi )

12

(8.19)

where

mi

isthe

mass

ofatomiandr

ithe

positionofatomi

with

respecttothe

centerofm

assof

them

olecule.Itis

especiallyusefulto

characterizepolym

ersolutions

andproteins.

Som

etimes

itis

interestingto

plotthedista

ncebetw

eentw

oatom

s,or

themin

imu

mdistance

be-tw

eentw

ogroups

ofatoms

(e.g.:protein

side-chainsin

asaltbridge).

Tocalculate

thesedistances

between

certaingroups

thereare

severalpossibilities:

•T

hed

istan

ceb

etw

ee

nth

ege

om

etrica

lcen

ters

oftwo

groupscan

becalculated

with

theprogram

gsg

an

gle

,asexplained

insec.8.6.

•T

hem

inim

um

dista

ncebetw

eentw

ogroups

ofatom

sduring

time

canbe

calculatedw

iththe

programg

min

dist

.Italso

calculatesthenu

mb

ero

fcon

tactsbetw

eenthese

groupsw

ithina

certainradiusr

max .

•To

monitorthem

inim

um

dista

nce

sb

etw

ee

nre

sidu

es

(seechapter5)w

ithina

(protein)molecule,

youcan

usethe

programgm

dm

at.

This

minim

umdistance

between

two

residuesA

iand

Aj

isdefined

asthe

smallestdistance

between

anypairofatom

s(i

∈A

i ,j∈A

j ).T

heoutput

isa

symm

etricalmatrix

ofsmallestdistances

between

allresidues.To

visualizethis

matrix,

youcan

usea

programsuch

asxv

.Ifyou

wantto

viewthe

axesand

legendor

ifyouw

antto

printthem

atrix,youcan

convertitwithxp

m2

ps

intoa

Postscriptpicture,see

Fig.8.7.

Plotting

thesem

atricesfor

differenttime-fram

es,one

cananalyze

changesin

thestructure,

ande.g.form

ingofsaltbridges.

Page 107: CS USER AL Simulations - GROMACS

14

2C

ha

pte

r8

.A

na

lysi

s

φ =

= 0

AB

Fig

ure

8.5:

Dih

edra

lcon

vent

ions

:A

.“B

ioch

emic

alco

nven

tion”

.B

.“P

olym

erco

nven

tion”

.

bb

a

φ

2

C

D

d

d

E

φ

d

φ

AB

n

1d

n

n

Fig

ure

8.6:

Opt

ions

ofgsg

an

gle

:A

.Ang

lebe

twee

n2

vect

ors.

B.A

ngle

betw

een

ave

ctor

and

the

norm

alof

apl

ane.

C.A

ngle

betw

een

two

plan

es.

D.D

ista

nce

betw

een

the

geom

etric

alce

nter

sof

2pl

anes

.E

.Dis

tanc

esbe

twee

na

vect

oran

dth

ece

nter

ofa

plan

e.

5.5

.D

ata

ba

ses

87

dihe

dral

s,be

caus

eth

ere

stis

read

from

the

coor

dina

tein

put

file

(inth

eca

seof

pd

b2

gm

x,

apd

bfo

rmat

file)

.S

ome

prot

eins

cont

ain

resi

dues

that

are

nots

tand

ard,

buta

relis

ted

inth

eco

ordi

nate

file.

You

have

toco

nstr

ucta

build

ing

bloc

kfo

rthi

s“s

tran

ge”r

esid

ue,o

ther

wis

eyo

uw

illno

tobt

ain

a*.

top

file.

Thi

sal

soho

lds

for

mol

ecul

esin

the

coor

dina

tefil

elik

eph

osph

ate

orsu

lpha

teio

ns.

The

resi

due

data

base

isco

nstr

ucte

din

the

follo

win

gw

ay:

[b

on

de

dty

pe

s]

;m

an

dato

ry;

bo

nd

sa

ng

les

dih

ed

rals

imp

rop

ers

11

12

;m

an

da

tory

[G

LY

];

ma

nd

ato

ry

[a

tom

s]

;m

an

da

tory

;n

am

ety

pe

cha

rge

cha

rge

gro

up

NN

-0.2

80

0H

H0

.28

00

CA

CH

20

.00

01

CC

0.3

80

2O

O-0

.38

02

[b

on

ds

];

op

tion

al

;ato

m1

ato

m2

b0

kbN

HN

CA

CA

CC

O-C

N

[e

xclu

sio

ns

];

op

tion

al

;ato

m1

ato

m2

[a

ng

les

];

op

tion

al

;ato

m1

ato

m2

ato

m3

th0

cth

[d

ihe

dra

ls]

;o

ptio

na

l;a

tom

1a

tom

2a

tom

3a

tom

4p

hi0

cpm

ult

[im

pro

pe

rs]

;o

ptio

na

l;a

tom

1a

tom

2a

tom

3a

tom

4q

0cq

N-C

CA

H-C

-CA

N-O

[Z

N]

[a

tom

s]

ZN

ZN

2.0

00

0

The

file

isfr

eefo

rmat

,th

eon

lyre

stric

tion

isth

atth

ere

can

beat

mos

ton

een

try

ona

line.

The

first

field

inth

efil

eis

the[

bo

nd

ed

typ

es

]fie

ld,

whi

chis

follo

wed

byfo

urnu

mbe

rs,

that

indi

cate

the

inte

ract

ion

type

for

bond

s,an

gles

,dih

edra

lsan

dim

prop

erdi

hedr

als.

The

file

cont

ains

resi

due

entr

ies,

whi

chco

nsis

tofa

tom

san

dop

tiona

llybo

nds,

angl

esdi

hedr

als

and

impr

oper

s.T

hech

arge

grou

pco

des

deno

teth

ech

arge

grou

pnu

mbe

rs.

Ato

ms

inth

esa

me

char

gegr

oup

shou

ld

Page 108: CS USER AL Simulations - GROMACS

88

Ch

ap

ter

5.

Top

olog

ies

always

bebelow

eachother.

When

usingthe

hydrogendatabase

with

pd

b2

gm

xforadding

missing

hydrogens,the

atomnam

esdefined

inthe.rtp

entryshould

correspondexactly

tothe

naming

conventionused

inthe

hydrogendatabase,

see5.5.2.

The

atomnam

esin

thebonded

interactioncan

bepreceded

bya

minus

ora

plus,indicating

thatthe

atomis

inthe

precedingor

following

residuerespectively.

Param

eterscan

beadded

tobonds,

angles,dihedrals

andim

propers,these

parameters

overridethe

standardparam

etersin

the.itp

files.T

hisshould

onlybe

usedin

specialcases.

Insteadof

parameters,

astring

canbe

addedfor

eachbonded

interaction,this

isused

inG

RO

MO

S96.rtp

files.T

hesestrings

arecopied

tothe

topologyfile

andcan

bereplaced

byforce

fieldparam

etersby

theC

-preprocessorin

gro

mp

pusing

#d

efin

estatem

ents.

pd

b2

gm

xautom

aticallygenerates

allangles,this

means

thatfor

theG

RO

MA

CS

forcefield

the[

an

gle

s]

fieldis

onlyusefulfor

overriding.itpparam

eters.F

orthe

GR

OM

OS

-96force

fieldthe

interactionnum

beroffallangles

needto

bespecified.

pd

b2

gm

xautom

aticallygenerates

oneproper

dihedralfor

everyrotatable

bond,preferably

onheavy

atoms.

When

the[d

ihe

dra

ls]

fieldis

used,noother

dihedralsw

illbegenerated

forthe

bondscorresponding

tothe

specifieddihedrals.

Itispossible

toputm

orethan

onedihedralon

arotatable

bond.

pd

b2

gm

xsets

thenum

berexclusions

to3,

which

means

thatinteractions

between

atoms

con-nected

byat

most

3bonds

areexcluded.

Pair

interactionsare

generatedfor

allpairs

ofatom

sw

hichare

seperatedby

3bonds

(exceptpairs

ofhydrogens).

When

more

interactionsneed

tobe

excluded,orsom

epair

interactionsshould

notbegenerated,an

[e

xclusio

ns

]field

canbe

added,follow

edby

pairsof

atomnam

eson

seperatelines.

Allnon-bonded

andpair

interactionsbetw

eenthese

atoms

willbe

excluded.

5.5.2H

ydrogendatabase

The

hydrogendatabase

isstored

inff?

??

.hd

b.

Itcontainsinform

ationfor

thepd

b2

gm

xpro-

gramon

howto

connecthydrogenatom

sto

existingatom

s.H

ydrogenatom

sare

named

afterthe

atomthey

areconnected

to:the

firstletteroftheatom

name

isreplaced

byan

’H’.Ifm

orethen

onehydrogen

atomis

connectedto

thesam

eatom

,anum

berw

illbeadded

tothe

endofthe

hydrogenatom

name.

For

example,adding

two

hydrogenatom

sto

ND

2(in

asparagine),thehydrogen

atoms

willbe

namedH

D2

1and

HD

22.

This

isim

portantsinceatom

naming

inthe.rtp

file(see5.5.1)

mustbe

thesam

e.T

heform

atofthehydrogen

databaseis

asfollow

s:

;re

s#

ad

ditio

ns

#H

ad

dtyp

ei

jk

AL

A11

1N

-CC

AA

RG

412

NC

AC

11

NE

CD

CZ

23

NH

1C

ZN

E2

3N

H2

CZ

NE

On

thefirstline

we

seethe

residuenam

e(A

LAor

AR

G)

andthe

number

ofadditions.A

fterthat

follows

oneline

foreach

addition,onw

hichw

esee:

•T

henum

berofH

atoms

added

8.6

.B

on

ds,a

ng

les

an

dd

ihe

dra

ls1

41

0.050.0 �

100.0 �

150.0 �

Tim

e (ps)

0.0

1000.0

2000.0

3000.0

4000.0

MSD (10-5 cm

2 s

-1)

Mean S

quare Displacem

entD

= 3.5027 (10

-5 cm2 s

-1)

Figure

8.4:M

eanS

quareD

isplacementofS

PC

-water.

Tom

onitorspecificb

on

dsin

yourm

oleculesduring

time,

theprogramg

bo

nd

calculatesthe

distributionof

thebond

lengthin

time.

The

indexfile

consistsof

pairsof

atomnum

bers,for

example

[b

on

ds_

1]

12

34

91

0[

bo

nd

s_2

]1

21

3

The

programg

an

gle

calculatesthe

distributionofan

gle

sandd

ihe

dra

lsintim

e.It

alsogives

theaverage

angleor

dihedral.T

heindex

fileconsists

oftripletsor

quadruplesofatom

numbers:

[a

ng

les

]1

23

23

43

45

[d

ihe

dra

ls]

12

34

23

55

For

thedihedralangles

youcan

useeither

the“biochem

icalconvention”(

φ=

0≡cis)

or“poly-

mer

convention”(φ

=0≡tra

ns),see

Fig.8.5.

Tofollow

specifican

gle

sintim

ebetw

eentw

ovectors,a

vectorand

aplane

ortw

oplanes

(definedby

2,resp.3

atoms

insideyour

molecule,see

Fig.

8.6A,B

,C),use

theprogramg

sga

ng

le.

Page 109: CS USER AL Simulations - GROMACS

14

0C

ha

pte

r8

.A

na

lysi

s

The

self

diffu

sion

coef

ficie

ntca

nbe

calc

ulat

edus

ing

the

Gre

en-K

ubo

rela

tion

[65

]

DA

=1 3

∫ ∞ 0〈v

i(t)·v

i(0)〉 i∈

Adt

(8.1

6)

whi

chis

just

the

inte

gral

ofth

eve

loci

tyau

toco

rrel

atio

nfu

nctio

n.T

here

isa

wid

ely

held

belie

fth

atth

eve

loci

tyA

CF

conv

erge

sfa

ster

than

the

mea

nsq

uare

disp

lace

men

t(se

c.8.

5.5)

,whi

chca

nal

sobe

used

for

the

com

puta

tion

ofdi

ffusi

onco

nsta

nts.

How

ever

,Alle

n&

Tild

esly

[65

]war

nus

that

the

long

time

cont

ribut

ion

toth

eve

loci

tyA

CF

can

notb

eig

nore

d,so

care

mus

tbe

take

n.

Ano

ther

impo

rtan

tqu

antit

yis

the

dipo

leco

rrel

atio

ntim

e.T

hedip

ole

corr

ela

tion

fun

ctio

nfor

part

icle

sAis

calc

ulat

edas

follo

ws

bygd

ipo

les

:

Cµ(τ

)=

〈µi(τ)·µ

i(0)〉 i∈

A(8

.17)

with

µi

=∑ j∈

ir jq j

.T

hedi

pole

corr

elat

ion

time

can

beco

mpu

ted

usin

geq

n.8.

8.F

orso

me

appl

icat

ions

see

[67].

The

visc

osity

ofa

liqui

dca

nbe

rela

ted

toth

eco

rrel

atio

ntim

eof

the

Pre

ssur

ete

nsor

P[6

8,69

].g

en

erg

yca

nco

mpu

teth

evi

scos

ity,b

utin

our

expe

rienc

eth

isis

notv

ery

accu

rate

(act

ually

the

valu

esdo

notc

onve

rge.

..).

8.5.

5M

ean

Squ

are

Dis

plac

emen

t

Tode

term

ine

the

self

diffu

sion

coef

ficie

ntDA

ofpa

rtic

lesA

one

can

use

the

Ein

stei

nre

latio

n[

65]

lim t→∞〈‖

r i(t

)−

r i(0

)‖2〉 i∈

A=

6DAt

(8.1

8)

Thi

sM

ea

nS

qu

are

Dis

pla

cem

en

tan

dD

Aar

eca

lcul

ated

byth

epr

ogra

mgm

sd.

For

mol

ecul

esco

nsis

ting

ofm

ore

than

one

atom

,r i

isth

ece

nter

ofm

ass

posi

tions

.In

that

case

you

shou

ldus

ean

inde

xfil

ew

ithm

olec

ule

num

bers

!T

hepr

ogra

mca

nal

sobe

used

for

calc

ulat

ing

diffu

sion

inon

eor

two

dim

ensi

ons.

Thi

sis

usef

ulfo

rst

udyi

ngla

tera

ldiff

usio

non

inte

rfac

es.

An

exam

ple

ofth

em

ean

squa

redi

spla

cem

ento

fSP

C-w

ater

isgi

ven

inF

ig.

8.4.

8.6

Bon

ds,a

ngle

san

ddi

hedr

als

g_

bo

nd

g_

an

gle

g_

sga

ng

le

5.5

.D

ata

ba

ses

89

•T

hew

ayof

addi

ngH

atom

s,ca

nbe

any

of

1o

ne

pla

na

rh

ydro

gen

,e.g

.rin

gs

or

pe

ptid

eb

on

don

ehy

drog

enat

om(n

)is

gene

rate

d,ly

ing

inth

epl

ane

ofat

oms

(i,j,k

)on

the

line

bise

ctin

gan

gle

(j-i-k

)at

adi

stan

ceof

0.1

nmfr

omat

omi,

such

that

the

angl

es(n

-i-j)

and

(n-i-

k)ar

e>90

degr

ees

2o

ne

sin

gle

hyd

roge

n,e

.g.h

ydro

xyl

one

hydr

ogen

atom

(n)i

sge

nera

ted

ata

dist

ance

of0.

1nm

from

atom

i,su

chth

atan

gle

(n-i-

j)=10

9.5

degr

ees

and

dihe

dral

(n-i-

j-k)=

tran

s

3tw

op

lan

ar

hyd

roge

ns,

e.g.

-NH 2

two

hydr

ogen

s(n

1,n2

)ar

ege

nera

ted

ata

dist

ance

of0.

1nm

from

atom

i,su

chth

atan

gle

(n1-

i-j)=

(n2-

i-j)=

120

degr

ees

and

dihe

dral

(n1-

i-j-k

)=ci

san

d(n

2-i-j

-k)=

tran

s,su

chth

atna

mes

are

acco

rdin

gto

IUP

AC

stan

dard

s[

61]

4tw

oo

rth

ree

tetr

ah

ed

ralh

ydro

gen

s,e.

g.-C

H3

thre

e(n

1,n2

,n3)

ortw

o(n

1,n2

)hy

drog

ens

are

gene

rate

dat

adi

stan

ceof

0.1

nmfr

omat

omi,

such

that

angl

e(n

1-i-j

)=(n

2-i-j

)=(n

3-i-j

)=10

9.5,

dihe

dral

(n1-

i-j-k

)=tr

ans,

(n2-

i-j-k

)=tr

ans+

120

and

(n3-

i-j-k

)=tr

ans+

240

degr

ees

5o

ne

tetr

ah

ed

ralh

ydro

gen

,e.g

.C 3CH

one

hydr

ogen

atom

(n1)

isge

nera

ted

ata

dist

ance

of0.

1nm

from

atom

iin

tetr

ahed

ral

conf

orm

atio

nsu

chth

atan

gle

(n1-

i-j)=

(n1-

i-k)=

(n1-

i-l)=

109.

5de

gree

s

6tw

ote

tra

he

dra

lhyd

roge

ns,

e.g.

C-C

H2-C

two

hydr

ogen

atom

s(n

1,n2

)ar

ege

nera

ted

ata

dist

ance

of0.

1nm

from

atom

iin

tetr

ahed

ralc

onfo

rmat

ion

onth

epl

ane

biss

ectin

gan

gle

i-j-k

with

angl

e(n

1-i-n

2)=

(n1-

i-j)=

(n1-

i-k)=

109.

5

7tw

ow

ate

rh

ydro

gen

stw

ohy

drog

ens

are

gene

rate

dar

ound

atom

iac

cord

ing

toS

PC

[48

]w

ater

geom

etry

.T

hesy

mm

etry

axis

will

alte

rnat

ebe

twee

nth

ree

coor

dina

teax

esin

both

dire

ctio

ns

•T

hree

orfo

urco

ntro

lato

ms

(i,j,k

,l),w

here

the

first

alw

ays

isth

eat

omto

whi

chth

eH

atom

sar

eco

nnec

ted.

The

othe

rtw

oor

thre

ede

pend

onth

eco

dese

lect

ed.

5.5.

3Te

rmin

idat

abas

e

The

term

inid

atab

ases

are

stor

edin

ff?

??

-n.td

ban

dff?

??

-c.td

bfo

rthe

N-a

ndC

-ter

min

ire

spec

tivel

y.T

hey

cont

ain

info

rmat

ion

fort

hepdb

2g

mx

prog

ram

onho

wto

conn

ectn

ewat

oms

toex

istin

gon

es,w

hich

atom

ssh

ould

bere

mov

edor

chan

ged

and

whi

chbo

nded

inte

ract

ions

shou

ldbe

adde

d.T

hefo

rmat

ofth

eis

asfo

llow

s(t

his

isan

exam

ple

from

the

ffg

mx-

c.td

b):

[N

on

e]

[C

OO

-]

[re

pla

ce]

CC

C1

2.0

11

0.2

7[

ad

d]

28

CC

AN

OO

M1

5.9

99

4-0

.63

5

Page 110: CS USER AL Simulations - GROMACS

90

Ch

ap

ter

5.

Top

olog

ies

[d

ele

te]

O[im

pro

pe

rs]

CO

1O

2C

A

The

fileis

organizedin

blocks,each

with

aheader

specifyingthe

name

ofthe

block.T

heseblocks

correspondto

differenttypes

ofterm

inithat

canbe

addedto

am

olecule.In

thisexam

-ple

[N

on

e]

isthe

firstblock,

correspondingto

aterm

inusthat

leavesthe

molecule

asit

is;[

CO

O-

]is

thesecond

terminus

type,correspondingto

changingthe

terminalcarbon

atominto

adeprotonated

carboxylgroup.B

locknam

escannot

beany

ofthe

following:

rep

lace

,a

dd

,d

ele

te,

bo

nd

s,

an

gle

s,

dih

ed

rals

,im

pro

pe

rs;

thisw

ouldinterfere

with

theparam

e-ters

oftheblock,and

would

probablyalso

bevery

confusingto

human

readers.

Per

blockthe

following

optionsare

present:

•[

rep

lace

]replace

anexisting

atomby

onew

itha

differentatomtype,atom

name,charge

and/orm

ass.F

oreach

atomto

bereplaced

online

shouldbe

enteredw

iththe

following

fields:

–nam

eofthe

atomto

bereplaced

–new

atomnam

e

–new

atomtype

–new

mass

–new

charge

•[

ad

d]

addnew

atoms.

For

each(group

of)added

atom(s),

atw

o-lineentry

isnecessary.

The

firstline

containsthe

same

fieldsas

anentry

inthe

hydrogendatabase

(number

ofatoms,type

ofaddition,

controlatoms,

see5.5.1),but

thepossible

typesof

additionare

extendedby

two

more,specifically

forC

-terminaladditions:

8tw

oca

rbo

xyloxyge

ns,-C

OO −

two

oxygens(n1,n2)

aregenerated

accordingto

rule3,ata

distanceof0.136

nmfrom

atomiand

anangle

(n1-i-j)=(n2-i-j)=

117degrees

9ca

rbo

xyloxyge

ns

an

dh

ydroge

n,-C

OO

Htw

ooxygens

(n1,n2)are

generatedaccording

torule

3,at

distancesof

0.123nm

and0.125

nmfrom

atomifor

n1and

n2resp.

andangles

(n1-i-j)=121

and(n2-i-j)=

115degrees.

One

hydrogen(n’)

isgenerated

aroundn2

accordingto

rule2,

where

n-i-jand

n-i-j-kshould

beread

asn’-n2-iand

n’-n2-i-jresp.

After

thisline

anotherline

follows

which

specifiesthe

detailsof

theadded

atom(s),

inthe

same

way

asfor

replacingatom

s,i.e.:

–atom

name

–atom

type

–m

ass

–charge

8.5

.C

orre

latio

nfu

nctio

ns

13

9

intervals(j<<M

),butitmakes

iteasierto

interprettheresults.

Another

aspectthatmay

notbeneglected

when

computing

AC

Fs

fromsim

ulation,isthatusually

thetim

eoriginsξ

(eqn.8.6)are

notstatisticallyindependent,w

hichm

ayintroduce

abias

inthe

results.T

hiscan

betested

usinga

block-averagingprocedure,w

hereonly

time

originsw

itha

spacingatleastthe

lengthofthe

time

lagare

included,e.g.usingktime

originsw

ithspacing

ofM∆t

(where

kM

≤N

):

Cf (j∆

t)=

1k

k−1

∑i=0

f(iM∆t)f((iM

+j)∆

t)(8.11)

How

ever,oneneeds

verylong

simulations

togetgood

accuracythis

way,because

thereare

many

fewer

pointsthatcontribute

tothe

AC

F.

8.5.2U

singF

FT

forcom

putationofthe

AC

F

The

computationalcostfor

calculatingan

AC

Faccording

toeqn.8.9

isproportionaltoN

2,which

isconsiderable.

How

ever,thiscan

beim

provedby

usingfastF

ouriertransform

sto

dothe

convo-lution

[65].

8.5.3S

pecialforms

oftheA

CF

There

aresom

eim

portantvarietieson

theA

CF,e.g.the

AC

Fofa

vectorp

:

Cp(t)

= ∫∞0P

n (cos6

(p(t),p(t+ξ))d

ξ(8.12)

where

Pn (x)

isthe

nth

orderLegendre

polynomial 1.

Such

correlationtim

escan

actuallybe

ob-tained

experimentally

usinge.g.N

MR

orother

relaxationexperim

ents.G

RO

MA

CS

cancom

putecorrelations

usingthe

1stand

2 nd

orderLegendre

polynomial(eqn.

8.12).T

hiscan

a.o.be

usedfor

rotationalautocorrelation(gro

tacf

),dipoleautocorrelation

(gd

ipo

les

).

Inorder

tostudy

torsionangle

dynamics

we

definea

dihedralautocorrelationfunction

as[

66]:

C(t)

=〈cos(θ(τ)−

θ(τ+t))〉

τ(8.13)

Note

thatthisis

notaproductoftw

ofunctions

asis

generallyused

forcorrelation

functions,butitm

aybe

rewritten

asthe

sumoftw

oproducts:

C(t)

=〈cos(θ(τ))cos(θ(τ

+t))

+sin(θ(τ))sin(θ(τ

+t))〉

τ(8.14)

8.5.4S

ome

Applications

The

programg

vela

cccalculates

thisVe

locity

Au

toC

orre

latio

nF

un

ction.

Cv(τ)

=〈v

i (τ)·v

i (0)〉i∈

A(8.15)

1P0 (x

)=

1,P1 (x

)=

x,P

2 (x)

=(3

x2−

1)/

2

Page 111: CS USER AL Simulations - GROMACS

13

8C

ha

pte

r8

.A

na

lysi

s

0.0

0.5

1.0

1.5

2.0

r (n

m)

0.0

1.0

2.0

3.0

g(r) �

Gro

mac

s R

DF

OW

1-O

W1

Fig

ure

8.3:g O

O(r

)fo

rO

xyge

n-O

xyge

nof

SP

C-w

ater

.

whe

reth

eno

tatio

non

the

right

hand

side

mea

nsav

erag

ing

over

ξ,i.e

.ove

rtim

eor

igin

s.It

isal

sopo

ssib

leto

com

pute

cros

s-co

rrel

atio

nfu

nctio

nfr

omtw

opr

oper

ties

f(t

)an

dg(t

):

Cfg(t

)=

〈f(ξ

)g(ξ

+t)〉 ξ

(8.7

)

how

ever

,in

GR

OM

AC

Sth

ere

isno

stan

dard

mec

hani

smto

doth

is(

note

:yo

uca

nus

eth

exmg

rpr

ogra

mto

com

pute

cros

sco

rrel

atio

ns).

The

inte

gral

ofth

eco

rrel

atio

nfu

nctio

nov

ertim

eis

the

corr

elat

ion

timeτ

f:

τ f=

∫ ∞ 0C

f(t

)dt

(8.8

)

Inpr

actic

eco

rrel

atio

nfu

nctio

nsar

eca

lcul

ated

base

don

data

poin

tsw

ithdi

scre

tetim

ein

terv

als

∆t,

soth

atth

eA

CF

from

anM

Dsi

mul

atio

nis

:

Cf(j

∆t)

=1

N−j

N−

1−

j ∑ i=0

f(i

∆t)f((i+j)

∆t)

(8.9

)

whe

reN

isth

enu

mbe

rofa

vaila

ble

time

fram

esfo

rthe

calc

ulat

ion.

The

resu

lting

AC

Fis

obvi

ousl

yon

lyav

aila

ble

attim

epo

ints

with

the

sam

ein

terv

al∆t.

Sin

cefo

rm

any

appl

icat

ions

itis

nece

ssar

yto

know

the

shor

ttim

ebe

havi

orof

the

AC

F(e

.g.t

hefir

st10

ps)

this

ofte

nm

eans

that

we

have

tosa

veth

eat

omic

coor

dina

tes

with

shor

tint

erva

ls.

Ano

ther

impl

icat

ion

ofeq

n.8.

9is

that

inpr

inci

ple

we

can

notc

ompu

teal

lpoi

nts

ofth

eA

CF

with

the

sam

eac

cura

cy,s

ince

we

have

N−

1da

tapo

ints

forC

f(∆t)

but

only

1fo

rCf((N−

1)∆t)

.H

owev

er,

ifw

ede

cide

toco

mpu

teon

lyan

AC

Fof

leng

thM

∆t,

whe

reM

≤N/2

we

can

com

pute

allp

oint

sw

ithth

esa

me

stat

istic

alac

cura

cy:

Cf(j

∆t)

=1 M

N−

1−

M ∑ i=0

f(i

∆t)f((i+j)

∆t)

(8.1

0)

here

ofco

ursej<M

.M

isso

met

imes

refe

rred

toas

the

time

lag

ofth

eco

rrel

atio

nfu

nctio

n.W

hen

we

deci

deto

doth

is,w

ein

tent

iona

llydo

notu

seal

lthe

avai

labl

epo

ints

for

very

shor

ttim

e

5.6

.F

ilefo

rma

ts9

1

Like

inth

ehy

drog

enda

taba

se(s

ee5.5.

1),

whe

nm

ore

then

one

atom

isco

nnec

ted

toan

exis

ting

one,

anu

mbe

rw

illbe

appe

nded

toth

een

dof

the

atom

nam

e.

•[

de

lete

]de

lete

exis

ting

atom

s.O

neat

omna

me

per

line.

•[

bo

nd

s]

,[a

ng

les

],[

dih

ed

rals

]an

d[

imp

rop

ers

]ad

dad

ditio

nalb

onde

dpa

ram

eter

s.T

hefo

rmat

isid

entic

alto

that

used

inth

eff?

??

.rtp

,se

e5.5

.1.

5.6

File

form

ats

5.6.

1To

polo

gyfil

e

The

topo

logy

file

isbu

iltfo

llow

ing

the

GR

OM

AC

Ssp

ecifi

catio

nfo

ra

mol

ecul

arto

polo

gy.

A*.

top

file

can

bege

nera

ted

bypd

b2

gm

x.

Des

crip

tion

ofth

efil

ela

yout

:

•se

mic

olon

(;)

and

new

line

surr

ound

com

men

ts

•on

alin

een

ding

with\

the

new

line

char

acte

ris

igno

red.

•di

rect

ives

are

surr

ound

edby[

and

]

•th

eto

polo

gyco

nsis

tsof

thre

ele

vels

:

–th

epa

ram

eter

leve

l(se

eTa

ble

5.3)

–th

em

olec

ule

leve

l,w

hich

shou

ldco

ntai

non

eor

mor

em

olec

ule

defin

ition

s(s

eeTa

-bl

e5.

4)

–th

esy

stem

leve

l:[sy

ste

m]

,[m

ole

cule

s]

•ite

ms

shou

ldbe

sepa

rate

dby

spac

esor

tabs

,not

com

mas

•at

oms

inm

olec

ules

shou

ldbe

num

bere

dco

nsec

utiv

ely

star

ting

at1

•th

efil

eis

pars

edon

ceon

lyw

hich

impl

ies

that

nofo

rwar

dre

fere

nces

can

betr

eate

d:ite

ms

mus

tbe

defin

edbe

fore

they

can

beus

ed

•ex

clus

ions

can

bege

nera

ted

from

the

bond

sor

over

ridde

nm

anua

lly

•th

ebo

nded

forc

ety

pes

can

bege

nera

ted

from

the

atom

type

sor

over

ridde

npe

rbo

nd

•de

scrip

tive

com

men

tlin

esan

dem

pty

lines

are

high

lyre

com

men

ded

•us

ing

one

ofth

e[a

tom

s]

,[

bo

nd

s]

,[

pa

irs

],

[a

ng

les

],

etc.

with

out

havi

ngus

ed[

mo

lecu

lety

pe

]be

fore

ism

eani

ngle

ssan

dge

nera

tes

aw

arni

ng.

•us

ing

[m

ole

cule

s]

with

out

havi

ngus

ed[

syst

em

]be

fore

ism

eani

ngle

ssan

dge

nera

tes

aw

arni

ng.

Page 112: CS USER AL Simulations - GROMACS

92

Ch

ap

ter

5.

Top

olog

ies

Param

etersinteraction

directive#

f.param

eterspert

typeat.

tp

ma

nd

ato

ryd

efa

ults

non-bondedfunction

type;com

binationrule;

generatepairs

(no/yes);fudge

LJ();fudge

QQ

()m

an

da

tory

ato

mtyp

es

atomtype;m

(u);q(e);particle

type;c6

(kJm

ol −1nm

6);c12

(kJm

ol −1nm

12)

bo

nd

type

s(see

Table5.4,directiveb

on

ds

)co

nstra

inttyp

es

(seeTable5.4,directive

con

strain

ts)

pa

irtype

s(see

Table5.4,directivep

airs

)a

ng

letyp

es

(seeTable5.4,directive

an

gle

s)

properdih.

dih

ed

raltyp

es

2(b

)1

θm

ax

(deg);fc(kJ

mol −

1);mult

X(a

)

improper

dih.d

ihe

dra

ltype

s2(c

)2

θ0

(deg);fc(kJ

mol −

1rad −2)

XR

Bdihedral

dih

ed

raltyp

es

2(b

)3

C0 ,C

1 ,C2 ,C

3 ,C4 ,C

5(kJ

mol −

1)LJ

no

nb

on

dp

ara

ms

21

c6(kJ

mol −

1nm6);c

12

(kJm

ol −1nm

12)

Buckingham

no

nb

on

dp

ara

ms

22

a(kJ

mol −

1);b(nm

−1);

c6

(kJm

ol −1nm

6)

Molecule

definition(s)

oneor

more

molecule

definitionsas

describedin

Table5.4

(nextpage)

System

ma

nd

ato

rysyste

msystem

name

ma

nd

ato

rym

ole

cule

sm

oleculenam

e;number

ofmolecules

’#at’is

thenum

berofatom

types’f.

tp’isfunction

type’pert’indicates

ifthisinteraction

typecan

beperturbed

duringfree

energycalculations

(a)

multiplicities

cannotbe

perturbed(b

)the

innertw

oatom

sin

thedihedral

(c)

theouter

two

atoms

inthe

dihedralF

orfree

energycalculations,the

parameters

fortopology

’B’(lam

bda=

1)should

beadded

onthe

same

line,afterthe

normalparam

eters,inthe

same

orderas

thenorm

alparameters.

Table5.3:

The

topology(*.to

p)

file.

8.5

.C

orre

latio

nfu

nctio

ns

13

7

r

r+dr

r+dr

rθ+

dθθ

e

ABD

CF

igure8.2:

Definition

ofslicesing

rdf

:A

.gA

B(r).

B.g

AB

(r,θ).T

heslices

arecolored

grey.C

.N

ormalization〈ρ

B 〉lo

cal .

D.N

ormalization〈ρ

B 〉lo

cal,

θ .N

ormalization

volumes

arecolored

grey.

Usually

thevalue

ofrm

ax

ishalf

ofthe

boxlength.

The

averagingis

alsoperform

edin

time.

Inpractice

theanalysis

programgrd

fdivides

thesystem

intosphericalslices

(fromrtor+dr,see

Fig.8.2A

)and

makes

ahistogram

instead

oftheδ-function.

An

example

ofthe

rdfof

Oxygen-

Oxygen

inS

PC

-water

[48]isgiven

inF

ig.8.3.

With

grd

fitis

alsopossible

tocalculate

anangle

dependentrdfgA

B(r,θ),w

herethe

angleθis

definedw

ithrespectto

acertain

laboratoryaxise,see

Fig.8.2B

.

gA

B(r,θ)

=1

〈ρB 〉

loca

l,θ

1NA

NA∑i∈

A

NB∑j∈

B

δ(rij −

r)δ(θij −

θ)2πr2sin(θ)

(8.4)

cos(θij )

=rij ·e

‖rij ‖

‖e‖(8.5)

This

gA

B(r,θ)

isusefulforanalyzing

anisotropicsystem

s.N

otethatin

thiscase

thenorm

alization〈ρ

B 〉lo

cal,

θis

theaverage

densityin

allangleslices

fromθ

toθ+

uptorm

ax ,so

angledependent,

seeF

ig.8.2D.

8.5C

orrelationfunctions

8.5.1T

heoryofcorrelation

functions

The

theoryofcorrelation

functionsis

wellestablished

[65].

How

everw

ew

anttodescribe

herethe

implem

entationofthe

variouscorrelation

functionflavors

inthe

GR

OM

AC

Scode.

The

definitionofthe

autocorrelationfunction

(AC

F)Cf (t)

fora

propertyf(t)is

Cf (t)

=〈f(ξ)f(ξ

+t)〉

ξ(8.6)

Page 113: CS USER AL Simulations - GROMACS

13

6C

ha

pte

r8

.A

na

lysi

s

Fig

ure

8.1:

The

win

dow

ofng

mx

show

ing

abo

xof

wat

er.

ase

tof

ener

gies

,lik

epo

tent

ial,

kine

ticor

tota

lene

rgy,

orin

divi

dual

cont

ribut

ions

,lik

eLe

nnar

d-Jo

nes

ordi

hedr

alen

ergi

es.

The

cen

ter-

of-

ma

ssve

loci

ty,defi

ned

as vco

m=

1 M

N ∑ i=1

miv

i(8

.2)

withM

=∑ N i=

1m

ith

eto

talm

ass

ofth

esy

stem

,can

bem

onito

red

intim

eby

the

prog

ram

gco

m.

Itis

how

ever

reco

mm

ende

dto

rem

ove

the

cent

er-o

f-m

ass

velo

city

ever

yst

ep(s

eech

apte

r3)

!

8.4

Rad

iald

istr

ibut

ion

func

tions

g_

rdf

The

rad

iald

istr

ibu

tion

fun

ctio

n(rd

f)or

pair

corr

elat

ion

func

tiong

AB

(r)

betw

een

part

icle

sof

type

Aan

dB

isde

fined

inth

efo

llow

ing

way

:

g AB

(r)

=〈ρ

B(r

)〉〈ρ

B〉 lo

cal

=1

〈ρB〉 lo

cal

1 NA

NA ∑ i∈A

NB ∑ j∈B

δ(r i

j−r)

4πr2

(8.3

)

with

〈ρB

(r)〉

the

part

icle

dens

ityof

typeB

ata

dist

ancer

arou

ndpa

rtic

lesA

,an

d〈ρ

B〉 lo

cal

the

part

icle

dens

ityof

typeB

aver

aged

over

all

sphe

res

arou

ndpa

rtic

les

Aw

ithra

diusr m

ax

(see

Fig

.8.2

C).

5.6

.F

ilefo

rma

ts9

3

Mol

ecul

ede

finiti

onin

tera

ctio

ndi

rect

ive

#f.

para

met

ers

pert

type

at.

tp

ma

nd

ato

rym

ole

cule

typ

em

olec

ule

nam

e;ex

clud

ene

ighb

ors

#bo

nds

away

for

non-

bond

edin

tera

ctio

nsm

an

da

tory

ato

ms

1at

omty

pe;r

esid

uenu

mbe

r;re

sidu

ena

me;

atom

nam

e;ch

arge

grou

pnu

mbe

r;q

(e);

m(u

)X

(b)

bond

bo

nd

s(c

,d)

21

b 0(n

m);

f c(k

Jm

ol−

1nm

−2)

XG

96bo

ndb

on

ds

(c,d

)2

2b 0

(nm

);f c

(kJ

mol−

1nm

−4)

Xm

orse

bo

nd

s(c

,d)

23

b 0(n

m);

D(k

Jm

ol−1);β

(nm−

1)

cubi

cbo

ndb

on

ds

(c,d

)2

4b 0

(nm

);C 2

(kJ

mol−

1nm

−2);

C3

(kJ

mol−

1nm

−3)

conn

ectio

nb

on

ds

(c)

25

harm

onic

pot.

bo

nd

s2

6b 0

(nm

);f c

(kJ

mol−

1nm

−2)

XLJ

1-4

pa

irs

21

c 6(k

Jm

ol−

1nm

6);

c 12

(kJ

mol−

1nm

12)

Xan

gle

an

gle

s(d

)3

1θ 0

(deg

);f c

(kJ

mol−

1ra

d−2)

XG

96an

gle

an

gle

s(d

)3

2θ 0

(deg

);f c

(kJ

mol−

1)

Xpr

oper

dih.

dih

ed

rals

41

θ max

(deg

);f c

(kJ

mol−

1);

mul

tX

(a)

impr

oper

dih.

dih

ed

rals

42

θ 0(d

eg);

f c(k

Jm

ol−

1ra

d−2)

XR

Bdi

hedr

ald

ihe

dra

ls4

3C 0

,C1,C

2,C

3,C

4,C

5(k

Jm

ol−

1)

cons

trai

ntco

nst

rain

ts2

1b 0

(nm

)X

cons

tr.n.

c.co

nst

rain

ts2

2b 0

(nm

)X

settl

ese

ttle

s3

1d O

H,d

HH

(nm

)du

mm

y2d

um

mie

s23

1a

()du

mm

y3d

um

mie

s34

1a,

b()

dum

my3

fdd

um

mie

s34

2a

();d

(nm

)du

mm

y3fa

dd

um

mie

s34

(deg

);d

(nm

)du

mm

y3ou

td

um

mie

s34

4a,

b()

;c(n

m−1)

dum

my4

fdd

um

mie

s45

1a,

b()

;d(n

m);

posi

tion

res.

po

sitio

nre

stra

ints

11

k x,k

y,k

z(k

Jm

ol−

1nm

−2)

dist

ance

res.

dis

tan

cere

stra

ints

21

type

;ind

ex;l

ow,u

p 1,up

2(n

m);

fact

or()

angl

ere

s.a

ng

lere

stra

ints

41

θ 0(d

eg);

f c(k

Jm

ol−

1);

mul

tX

(a)

angl

ere

s.z

an

gle

rest

rain

tsz

21

θ 0(d

eg);

f c(k

Jm

ol−

1);

mul

tX

(a)

excl

usio

nse

xclu

sio

ns

1on

eor

mor

eat

omin

dice

s

’#at

’is

the

num

ber

ofat

omin

dice

s’f.

tp’i

sfu

nctio

nty

pe’p

ert’

indi

cate

sif

this

inte

ract

ion

type

can

bepe

rtur

bed

durin

gfr

eeen

ergy

calc

ulat

ions

(a)

mul

tiplic

ities

can

notb

epe

rtur

bed

(b)

only

the

atom

type

,cha

rge

and

mas

sca

nbe

pert

urbe

d(c

)us

edby

gro

mp

pfo

rge

nera

ting

excl

usio

ns(d

)ca

nbe

conv

erte

dto

cons

trai

nts

bygro

mp

pF

orfr

eeen

ergy

calc

ulat

ions

,the

para

met

ers

for

topo

logy

’B’(

lam

bda

=1)

shou

ldbe

adde

don

the

sam

elin

e,af

ter

the

norm

alpa

ram

eter

s,in

the

sam

eor

der

asth

eno

rmal

para

met

ers.

Tabl

e5.

4:T

hem

olec

ule

defin

ition

.

Page 114: CS USER AL Simulations - GROMACS

94

Ch

ap

ter

5.

Top

olog

ies

•after[

system

]the

onlyallow

eddirective

is[m

ole

cule

s]

•using

anunknow

nstring

in[]

causesallthe

datauntilthe

nextdirective

tobe

ignored,and

generatesa

warning.

Here

isan

example

ofatopology

file,u

rea

.top

:

;;E

xam

ple

top

olo

gy

file;;

Th

efo

rcefie

ldfile

sto

be

inclu

de

d#

inclu

de

"ffgm

x.itp"

[m

ole

cule

type

];

na

me

nre

xclU

rea

3

[a

tom

s]

;n

rtyp

ere

snr

resid

ua

tom

cgn

rch

arg

e1

C1

UR

EA

C1

10

.68

32

O1

UR

EA

O2

1-0

.68

33

NT

1U

RE

AN

32

-0.6

22

4H

1U

RE

AH

42

0.3

46

5H

1U

RE

AH

52

0.2

76

6N

T1

UR

EA

N6

3-0

.62

27

H1

UR

EA

H7

30

.34

68

H1

UR

EA

H8

30

.27

6

[b

on

ds

];

ai

aj

fun

ctb

0kb

34

11

.00

00

00

e-0

13

.74

46

80

e+

05

35

11

.00

00

00

e-0

13

.74

46

80

e+

05

67

11

.00

00

00

e-0

13

.74

46

80

e+

05

68

11

.00

00

00

e-0

13

.74

46

80

e+

05

12

11

.23

00

00

e-0

15

.02

08

00

e+

05

13

11

.33

00

00

e-0

13

.76

56

00

e+

05

16

11

.33

00

00

e-0

13

.76

56

00

e+

05

[p

airs

];

ai

aj

fun

ctc6

c12

24

10

.00

00

00

e+

00

0.0

00

00

0e

+0

02

51

0.0

00

00

0e

+0

00

.00

00

00

e+

00

27

10

.00

00

00

e+

00

0.0

00

00

0e

+0

02

81

0.0

00

00

0e

+0

00

.00

00

00

e+

00

37

10

.00

00

00

e+

00

0.0

00

00

0e

+0

03

81

0.0

00

00

0e

+0

00

.00

00

00

e+

00

46

10

.00

00

00

e+

00

0.0

00

00

0e

+0

05

61

0.0

00

00

0e

+0

00

.00

00

00

e+

00

[a

ng

les

];

ai

aj

ak

fun

ctth

0cth

13

41

1.2

00

000

e+

02

2.9

28

80

0e

+0

2

8.2

.L

oo

king

atyo

ur

traje

ctory

13

5

Sid

eC

ha

inprotein

sidechain

atoms;

thatis

allatoms

exceptN

,C

α ,C

,O

,backbone

amide

hydrogen,oxygens

inC

-terminus

andhydrogens

onthe

N-term

inus

Sid

eC

ha

in-H

proteinside

chainatom

sexcluding

allhydrogens

Pro

t-Ma

sses

proteinatom

sexcluding

dumm

ym

asses(as

usedin

dumm

yatom

constructionsof

NH

3

groupsand

Tryptophane

sidechains),seealso

sec.5.2.2;this

groupis

onlyincluded

when

itdiffers

fromthe

’Pro

tein

’group

No

n-P

rote

inallnon-protein

atoms

DN

AallD

NA

atoms

mo

lecu

len

am

efor

allresidues/m

oleculesw

hichare

notrecognized

asprotein

orD

NA

,one

groupper

residue/molecule

name

isgenerated

Oth

erallatom

sw

hichare

neitherprotein

norD

NA

.

Em

ptygroups

willnotbe

generated.

8.2Looking

atyourtrajectory

ng

mx

Before

analyzingyour

trajectoryit

isoften

informative

tolook

atyour

trajectoryfirst.

Grom

acscom

esw

itha

simple

trajectoryview

ern

gm

x;theadvantage

with

thisone

isthatitdoes

notrequireO

penGL,w

hichusually

isn’tpresente.g.on

supercomputers.

Itisalso

possibleto

generatea

hard-copy

inE

ncapsulatedP

ostscriptformat,

seeF

ig.8.1.

Ifyouw

antafaster

andm

orefancy

viewer

thereare

severalprograms

thatcan

readthe

GR

OM

AC

Strajectory

formats

–have

alook

atour

homepagew

ww

.gromacs.orgfor

updatedlinks.

8.3G

eneralproperties

g_

en

erg

yg

_co

m

Toanalyze

some

orallen

erg

iesand

otherproperties,

suchastota

lp

ressu

re,p

ressu

rete

nso

r,d

en

sity,bo

x-volu

meand

bo

x-sizes,use

theprogramg

en

erg

y.

Achoice

canbe

made

froma

list

Page 115: CS USER AL Simulations - GROMACS

13

4C

ha

pte

r8

.A

na

lysi

s

Gro

ups

can

ther

efor

eco

nsis

tof

ase

ries

ofa

tom

nu

mb

ers,

but

inso

me

case

sal

soofm

ole

cule

nu

mb

ers.

Itis

also

poss

ible

tosp

ecify

ase

ries

ofan

gles

bytrip

les

ofa

tom

nu

mb

ers,

dihe

dral

sby

qu

ad

rup

leso

fa

tom

nu

mb

ersa

ndbo

nds

orve

ctor

s(in

am

olec

ule)

bypairs

ofa

tom

nu

mb

ers.

Whe

nap

prop

riate

the

type

ofin

dex

file

will

besp

ecifi

edfo

rth

efo

llow

ing

anal

ysis

prog

ram

s.To

help

crea

ting

such

inde

xfil

es(

ind

ex.

nd

x),

ther

ear

ea

coup

leof

prog

ram

sto

gene

rate

them

,us

ing

eith

eryo

urin

put

confi

gura

tion

orth

eto

polo

gy.

Toge

nera

tean

inde

xfil

eco

nsis

ting

ofa

serie

sof

ato

mn

um

be

rs(as

inth

eex

ampl

eofg

AB

)us

ema

ken

dx

.To

gene

rate

anin

dex

file

with

angl

esor

dihe

dral

s,us

emk

an

gn

dx

.O

fco

urse

you

can

also

mak

eth

emby

hand

.T

hege

nera

lfo

rmat

ispr

esen

ted

here

:

[O

xyg

en

]1

47

[H

ydro

ge

n]

23

56

89

Firs

tth

egr

oup

nam

eis

writ

ten

betw

een

squa

rebr

acke

ts.

The

follo

win

gat

omnu

mbe

rsm

aybe

spre

adou

tove

ras

man

ylin

esas

you

like.

The

atom

num

berin

gst

arts

at1.

8.1.

1D

efau

ltG

roup

s

Whe

nno

inde

xfil

eis

supp

lied

toan

alys

isto

ols,

anu

mbe

rof

defa

ult

grou

psca

nbe

gene

rate

dto

choo

sefr

om:

Sys

tem al

lato

ms

inth

esy

stem

Pro

tein al

lpro

tein

atom

s

Pro

tein

-Hpr

otei

nat

oms

excl

udin

ghy

drog

ens

C-a

lph

a Cα

atom

s

Ba

ckb

on

epr

otei

nba

ckbo

neat

oms;

N,C α

and

C

Ma

inC

ha

inpr

otei

nm

ain

chai

nat

oms:

N,C α

,Can

dO

,inc

ludi

ngox

ygen

sin

C-t

erm

inus

Ma

inC

ha

in+

Cb

prot

ein

mai

nch

ain

atom

sin

clud

ing

Ma

inC

ha

in+

Hpr

otei

nm

ain

chai

nat

oms

incl

udin

gba

ckbo

neam

ide

hydr

ogen

and

hydr

ogen

son

the

N-

term

inus

5.6

.F

ilefo

rma

ts9

5

13

51

1.2

00

00

0e

+0

22

.92

88

00

e+

02

43

51

1.2

00

00

0e

+0

23

.34

72

00

e+

02

16

71

1.2

00

00

0e

+0

22

.92

88

00

e+

02

16

81

1.2

00

00

0e

+0

22

.92

88

00

e+

02

76

81

1.2

00

00

0e

+0

23

.34

72

00

e+

02

21

31

1.2

15

00

0e

+0

25

.02

08

00

e+

02

21

61

1.2

15

00

0e

+0

25

.02

08

00

e+

02

31

61

1.1

70

00

0e

+0

25

.02

08

00

e+

02

[d

ihe

dra

ls]

;a

ia

ja

ka

lfu

nct

ph

icp

mu

lt2

13

41

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

06

13

41

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

02

13

51

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

06

13

51

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

02

16

71

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

03

16

71

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

02

16

81

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

03

16

81

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

0

[d

ihe

dra

ls]

;a

ia

ja

ka

lfu

nct

q0

cq3

45

12

0.0

00

00

0e

+0

01

.67

36

00

e+

02

67

81

20

.00

00

00

e+

00

1.6

73

60

0e

+0

21

36

22

0.0

00

00

0e

+0

01

.67

36

00

e+

02

[p

osi

tion

_re

stra

ints

];

you

wo

uld

n’t

no

rma

llyu

seth

isfo

ra

mo

lecu

lelik

eU

rea

,;

bu

tit’

sh

ere

for

did

act

ica

lp

urp

ose

s;

ai

fun

ctfc

11

10

00

10

00

10

00

;R

est

rain

toa

po

int

21

10

00

01

00

0;

Re

stra

into

alin

e(Y

-axi

s)3

11

00

00

0;

Re

stra

into

ap

lan

e(Y

-Z-p

lan

e)

;In

clu

de

SP

Cw

ate

rto

polo

gy

#in

clu

de

"sp

c.itp

"

[sy

ste

m]

Ure

ain

Wa

ter

[m

ole

cule

s]

;mo

lecu

len

am

en

r.U

rea

1S

OL

10

00

Her

efo

llow

sth

eex

plan

ator

yte

xt.

[d

efa

ults

]:

•no

n-bo

ndty

pe=

1(L

enna

rd-J

ones

)or

2(B

ucki

ngha

m)

note

:w

hen

usin

gth

eB

ucki

ngha

mpo

tent

ial

noco

mbi

natio

nru

leca

nbe

used

,an

da

full

inte

ract

ion

mat

rixm

ustb

epr

ovid

edun

der

the

no

nb

on

dp

ara

ms

sect

ion.

Page 116: CS USER AL Simulations - GROMACS

96

Ch

ap

ter

5.

Top

olog

ies

•com

binationrule

=1

(supplyC(6

)and

C(1

2),σ

ij=√σ

i σj ),

2(supply

σand

ε,σ

ij=

12 (σi +

σj ))

or3

(supplyσand

ε,σij

=√σ

i σj )

•generate

pairs=

no(get1-4

interactionsfrom

pairlist)oryes(generate

1-4interactions

fromnorm

alLennard-Jonesparam

etersusing

FudgeLJ

andF

udgeQQ

)

•F

udgeLJ=

factorto

changeLennard-Jones

1-4interactions

•F

udgeQQ

=factor

tochange

electrostatic1-4

interactions

note:FudgeLJ

andF

udgeQQ

onlyneed

tobe

specifiedw

hengenerate

pairsis

setto’yes’.

#in

clud

e"ffg

mx.itp

":

thisincludes

thebonded

andnon-bonded

GR

OM

AC

Sparam

eters,the

gm

xin

ffgm

xw

illbereplaced

bythe

name

oftheforce

fieldyou

areactually

using.

[m

ole

cule

type

]:

definesthe

name

ofyourm

oleculein

this*.to

pand

nrexcl=3

standsfor

excludingnon-bonded

interactionsbetw

eenatom

sthatare

nofurther

than3

bondsaw

ay.

[a

tom

s]

:defines

them

olecule,w

herenrand

type

arefixed,

therest

isuser

defined.S

oa

tom

canbe

named

asyou

like,cgnr

made

largeror

smaller

(ifpossible,

thetotalcharge

ofa

chargegroup

shouldbe

zero),andcharges

canbe

changedhere

too.

[b

on

ds

]:

nocom

ment.

[p

airs

]:

1-4interactions

[a

ng

les

]:

nocom

ment

[d

ihe

dra

ls]

:in

thiscase

thereare

9proper

dihedrals(funct

=1),

3im

proper(funct

=2)

andno

Ryckaert-B

ellemans

typedihedrals.

Ifyou

want

toinclude

Ryckaert-B

ellemans

typedihedrals

ina

topology,dothe

following

(incase

ofe.g.decane):

[d

ihe

dra

ls]

;a

ia

ja

ka

lfu

nct

c0c1

c21

23

43

23

45

3

anddo

notforgettoera

seth

e1

-4in

tera

ctionin

[p

airs

]!!

[p

ositio

nre

strain

ts]

:harm

onicallyrestrain

theselected

particlesto

referenceposi-

tions(sec.4.2.9).

The

referencepositions

areread

froma

separatecoordinate

fileby

grompp.

#in

clud

e"sp

c.itp"

:includes

atopology

filethat

was

alreadyconstructed

(seenext

sec-tion,m

olecule.itp).

[syste

m]

:title

ofyoursystem

,userdefined

[m

ole

cule

s]

:this

definesthe

totalnumber

of(sub)m

oleculesin

yoursystem

thatare

de-fined

inthis*.to

p.

Inthis

example

fileit

standsfor

1urea

molecules

dissolvedin

1000w

aterm

olecules.T

hem

oleculetype

SO

Lis

definedin

thesp

c.itpfile.

5.6.2M

olecule.itpfile

Ifyou

constructa

topologyfile

youw

illusem

oreoften

(likea

water

molecule,

spc.itp

)it

isbetter

tom

akeam

ole

cule

.itpfile,w

hichonly

liststhe

information

ofthem

olecule:

Chapter

8

Analysis

Inthis

chapterdifferentw

aysofanalyzing

yourtrajectory

aredescribed.

The

names

ofthecorre-

spondinganalysis

programs

aregiven.

Specific

infoon

thein-and

outputoftheseprogram

scan

befound

inthe

on-linem

anualatwww

.gromacs.org.

The

outputfilesare

oftenproduced

asfinished

Grace/X

mgr

graphs.

First

insec.8.1

thegroup

conceptin

analysisis

explained.T

henthe

differentanalysis

toolsare

presented.

8.1G

roupsin

Analysis.

ma

ke_

nd

xm

k_a

ng

nd

x

Inchapter3

itw

asexplained

howgro

up

so

fa

tom

scanbe

usedin

theM

D-program

.In

most

analysisprogram

sgroups

ofatom

sare

neededto

work

on.M

ostprogram

scan

generateseveral

defaultindexgroups,butgroups

canalw

aysbe

readfrom

anindex

file.Let’s

considerasim

ulationof

abinary

mixture

ofcom

ponentsA

andB

.W

henw

ew

antto

calculatethe

radialdistribution

function(rdf)g

AB

(r)ofA

with

respecttoB

,we

haveto

calculate

4πr2g

AB

(r)=

VN

A∑i∈

A

NB∑j∈

B

P(r)

(8.1)

whereV

isthe

volume

andP(r)

isthe

probabilityto

finda

Batom

atadistance

rfrom

anA

atom.

By

havingthe

userdefine

theatom

nu

mb

ersfor

groupsA

andB

ina

simple

filew

ecan

calculatethis

gA

Bin

them

ostgeneral

way,

without

havingto

make

anyassum

ptionsin

therdf-program

aboutthetype

ofparticles.

Page 117: CS USER AL Simulations - GROMACS

13

2C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

dods

spas

sign

sse

cond

ary

stru

ctur

ean

dca

lcul

ates

solv

enta

cces

sibl

esu

rfac

ear

eag

chi

calc

ulat

esev

eryt

hing

you

wan

tto

know

abou

tchi

and

othe

rdi

hedr

als

ghe

lixca

lcul

ates

ever

ythi

ngyo

uw

antt

okn

owab

outh

elic

esg

ram

aco

mpu

tes

Ram

acha

ndra

npl

ots

xram

ash

ows

anim

ated

Ram

acha

ndra

npl

ots

whe

elpl

ots

helic

alw

heel

s

Inte

rfac

esg

pote

ntia

lca

lcul

ates

the

elec

tros

tatic

pote

ntia

lacr

oss

the

box

gde

nsity

calc

ulat

esth

ede

nsity

ofth

esy

stem

gor

der

com

pute

sth

eor

der

para

met

erpe

rat

omfo

rca

rbon

tails

gh2

orde

rco

mpu

tes

the

orie

ntat

ion

ofw

ater

mol

ecul

esg

bund

lean

alyz

esbu

ndle

sof

axes

,e.g

.tra

nsm

embr

ane

helic

es

Cov

aria

nce

anal

ysis

gco

var

calc

ulat

esan

ddi

agon

aliz

esth

eco

varia

nce

mat

rixg

anae

igan

alyz

esth

eei

genv

ecto

rs

Nor

mal

mod

esgr

ompp

mak

esa

run

inpu

tfile

mdr

unfin

dsa

pote

ntia

lene

rgy

min

imum

nmru

nca

lcul

ates

the

Hes

sian

gnm

eig

diag

onal

izes

the

Hes

sian

gan

aeig

anal

yzes

the

norm

alm

odes

gnm

ens

gene

rate

san

ense

mbl

eof

stru

ctur

esfr

omth

eno

rmal

mod

es

5.6

.F

ilefo

rma

ts9

7

[m

ole

cule

typ

e]

;n

am

en

rexc

lU

rea

3

[a

tom

s]

;n

rty

pe

resn

rre

sid

ua

tom

cgn

rch

arg

e1

C1

UR

EA

C1

10

.68

3..

..........

.....

............

.....

8H

1U

RE

AH

83

0.2

76

[b

on

ds

];

ai

aj

fun

ctc0

c13

41

1.0

00

00

0e

-01

3.7

44

68

0e

+0

5..

..........

.....

............

.....

16

11

.33

00

00

e-0

13

.76

56

00

e+

05

[p

airs

];

ai

aj

fun

ctc0

c12

41

0.0

00

00

0e

+00

0.0

00

00

0e

+0

0..

..........

.....

............

.....

56

10

.00

00

00

e+

00

0.0

00

00

0e

+0

0

[a

ng

les

];

ai

aj

ak

fun

ctc0

c11

34

11

.20

00

00

e+

02

2.9

28

80

0e

+0

2..

..........

.....

............

.....

31

61

1.1

700

00

e+

02

5.0

20

80

0e

+0

2

[d

ihe

dra

ls]

;a

ia

ja

ka

lfu

nct

c0c1

c22

13

41

1.8

00

00

0e

+0

23

.34

72

00

e+

01

2.0

00

00

0e

+0

0..

..........

.....

............

.....

31

68

11

.80

00

00

e+

02

3.3

47

20

0e

+0

12

.00

00

00

e+

00

[d

ihe

dra

ls]

;a

ia

ja

ka

lfu

nct

c0c1

34

51

20

.00

00

00

e+

00

1.6

73

60

0e

+0

26

78

12

0.0

00

00

0e

+0

01

.67

36

00

e+

02

13

62

20

.00

00

00

e+

00

1.6

73

60

0e

+0

2

Thi

sre

sults

ina

very

shor

t*.to

pfil

eas

desc

ribed

inth

epr

evio

usse

ctio

n,bu

tthi

stim

eyo

uon

lyne

edto

incl

ude

files

:

;T

he

forc

efie

ldfil

es

tob

ein

clu

de

d#

incl

ud

e"f

fgm

x.itp

"

;In

clu

de

ure

ato

po

log

y

Page 118: CS USER AL Simulations - GROMACS

98

Ch

ap

ter

5.

Top

olog

ies

#in

clud

e"u

rea

.itp"

;In

clud

eS

PC

wa

ter

top

olo

gy

#in

clud

e"sp

c.itp"

[syste

m]

Ure

ain

Wa

ter

[m

ole

cule

s]

;mo

lecu

len

am

en

um

be

rU

rea

1S

OL

10

00

5.6.3Ifdefoption

Avery

powerful

featurein

GR

OM

AC

Sis

theuse

of#

ifde

fstatem

entsin

your*.top

file.B

ym

akinguse

ofthis

statement,

differentparam

etersfor

onem

oleculecan

beused

inthe

same

*.top

file.A

nexam

pleis

givenfor

TF

E,w

herethere

isan

optionto

usedifferentcharges

onthe

atoms:

chargesderived

byD

eLoof

eta

l.[62]orby

VanB

uurenand

Berendsen

[43].

Infactyou

canuse

alltheoptions

oftheC

-Preprocessor,

cpp

,becausethis

isused

toscan

thefile.

The

way

tom

akeuse

ofthe#ifd

ef

optionis

asfollow

s:

•in

gro

mp

p.m

dp

(theG

RO

MA

CS

preprocessorinputparam

eters)use

theoption

de

fine

=-D

De

loo

ford

efin

e=

•putthe#

ifde

fstatem

entsin

your*.top

,asshow

nbelow

:

...

[a

tom

s]

;n

rtyp

ere

snr

resid

ua

tom

cgn

rch

arg

em

ass

#ifd

ef

De

Lo

of

;U

seC

ha

rge

sfro

mD

eL

oo

f1

C1

TF

EC

10

.74

2F

1T

FE

F1

-0.2

53

F1

TF

EF

1-0

.25

4F

1T

FE

F1

-0.2

55

CH

21

TF

EC

H2

10

.25

6O

A1

TF

EO

A1

-0.6

57

HO

1T

FE

HO

10

.41

#e

lse;

Use

Ch

arg

es

from

Va

nB

uu

ren

1C

1T

FE

C1

0.5

92

F1

TF

EF

1-0

.23

F1

TF

EF

1-0

.24

F1

TF

EF

1-0

.25

CH

21

TF

EC

H2

10

.26

6O

A1

TF

EO

A1

-0.5

5

7.4

.P

rogra

ms

by

top

ic1

31

Distances

instructures

overtim

e

gm

indistcalculates

them

inimum

distancebetw

eentw

ogroups

gdist

calculatesthe

distancesbetw

eenthe

centersofm

assoftw

ogroups

gm

dmat

calculatesresidue

contactmaps

grm

sdistcalculates

atompair

distancesaveraged

with

power

2,-3or

-6

Mass

distributionproperties

overtim

e

gtraj

plotsx,v,f,box,tem

peratureand

rotationalenergyg

gyratecalculates

theradius

ofgyrationg

msd

calculatesm

eansquare

displacements

grotacf

calculatesthe

rotationalcorrelationfunction

form

oleculesg

rdfcalculates

radialdistributionfunctions

Analyzing

bondedinteractions

gbond

calculatesbond

lengthdistributions

mk

angndxgenerates

indexfiles

forgangle

gangle

calculatesdistributions

andcorrelations

forangles

anddihedrals

gdih

analyzesdihedraltransitions

Structuralproperties

ghbond

computes

andanalyzes

hydrogenbonds

gsaltbr

computes

saltbridgesg

sascom

putessolventaccessible

surfacearea

gorder

computes

theorder

parameter

peratom

forcarbon

tailsg

sganglecom

putesthe

angleand

distancebetw

eentw

ogroups

gsorient

analyzessolventorientation

aroundsolutes

gbundle

analyzesbundles

ofaxes,e.g.helicesg

disreanalyzes

distancerestraints

Kinetic

properties

gtraj

plotsx,v,f,box,tem

peratureand

rotationalenergyg

velacccalculates

velocityautocorrelation

functionsg

tcafcalculates

viscositiesofliquids

Electrostatic

properties

geniongenerates

mono

atomic

ionson

energeticallyfavorable

positionsg

potentialcalculates

theelectrostatic

potentialacrossthe

boxg

dipolescom

putesthe

totaldipoleplus

fluctuationsg

dielectriccalculates

frequencydependentdielectric

constants

Protein

specificanalysis

Page 119: CS USER AL Simulations - GROMACS

13

0C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

geni

onge

nera

tes

mon

oat

omic

ions

onen

erge

tical

lyfa

vora

ble

posi

tions

genc

onf

mul

tiplie

sa

conf

orm

atio

nin

’rand

om’o

rient

atio

nsge

npr

gene

rate

spo

sitio

nre

stra

ints

for

inde

xgr

oups

prot

onat

epr

oton

ates

stru

ctur

es

Run

ning

asi

mul

atio

ngr

ompp

mak

esa

run

inpu

tfile

tpbc

onv

mak

esa

run

inpu

tfile

for

rest

artin

ga

cras

hed

run

mdr

unpe

rfor

ms

asi

mul

atio

nxm

drun

perf

orm

ssi

mul

atio

nsw

ithex

tra

expe

rimen

talf

eatu

res

Vie

win

gtr

ajec

torie

sng

mx

disp

lays

atr

ajec

tory

trjc

onv

conv

erts

traj

ecto

ries

toe.

g.pd

bw

hich

can

bevi

ewed

with

e.g.

rasm

ol

Pro

cess

ing

ener

gies

gen

ergy

writ

esen

ergi

esto

xvg

files

and

disp

lays

aver

ages

gen

emat

extr

acts

anen

ergy

mat

rixfr

oman

ener

gyfil

em

drun

with

-rer

un(r

e)ca

lcul

ates

ener

gies

for

traj

ecto

ryfr

ames

Con

vert

ing

files

editc

onf

conv

erts

and

man

ipul

ates

stru

ctur

efil

estr

jcon

vco

nver

tsan

dm

anip

ulat

estr

ajec

tory

files

trjc

atco

ncat

enat

estr

ajec

tory

files

enec

onv

conv

erts

ener

gyfil

esxm

p2ps

conv

erts

XP

Mm

atric

esto

enca

psul

ated

post

scrip

t(or

XP

M)

Tool

s mak

end

xm

akes

inde

xfil

esm

kan

gndx

gene

rate

sin

dex

files

for

gangl

egm

xche

ckch

ecks

and

com

pare

sfil

esgm

xdum

pm

akes

bina

ryfil

eshu

man

read

able

gtr

ajpl

ots

x,v

and

fofs

elec

ted

atom

s/gr

oups

(and

mor

e)fr

oma

traj

ecto

ryg

anal

yze

anal

yzes

data

sets

trjo

rder

orde

rsm

olec

ules

acco

rdin

gto

thei

rdi

stan

ceto

agr

oup

Dis

tanc

esbe

twee

nst

ruct

ures

grm

sca

lcul

ates

rmsd

’sw

itha

refe

renc

est

ruct

ure

and

rmsd

mat

rices

gco

nfrm

sfit

stw

ost

ruct

ures

and

calc

ulat

esth

erm

sdg

clus

ter

clus

ters

stru

ctur

esg

rmsf

calc

ulat

esat

omic

fluct

uatio

ns

5.6

.F

ilefo

rma

ts9

9

7H

O1

TF

EH

O1

0.3

#e

nd

if

[b

on

ds

];

ai

aj

fun

ctc0

c16

71

1.0

00

00

0e

-01

3.1

38

00

0e

+0

51

21

1.3

60

00

0e

-01

4.1

84

00

0e

+0

51

31

1.3

60

00

0e

-01

4.1

84

00

0e

+0

51

41

1.3

60

00

0e

-01

4.1

84

00

0e

+0

51

51

1.5

30

00

0e

-01

3.3

47

00

0e

+0

55

61

1.4

30

00

0e

-01

3.3

47

00

0e

+0

5

... 5.6.

4F

ree

ener

gyca

lcul

atio

ns

Fre

een

ergy

diffe

renc

esbe

twee

ntw

osy

stem

sA

and

Bca

nbe

calc

ulat

edas

desc

ribed

inse

c.3.

12.

The

syst

ems

Aan

dB

are

desc

ribed

byto

polo

gies

cons

istin

gof

the

sam

enu

mbe

rof

mol

ecul

esw

ithth

esa

me

num

ber

ofat

oms.

Mas

ses

and

non-

bond

edin

tera

ctio

nsca

nbe

pert

urbe

dby

addi

ngB

para

met

ers

inth

e[a

tom

s]

field

.B

onde

din

tera

ctio

nsca

nbe

pert

urbe

dby

addi

ngB

pa-

ram

eter

sto

the

bond

edty

pes

orth

ebo

nded

inte

ract

ions

.T

hepa

ram

eter

sth

atca

nbe

pert

urbe

dar

elis

ted

inTa

ble5

.3an

dTa

ble5

.4.

Theλ

-dep

ende

nce

ofth

ein

tera

ctio

nsis

desc

ribed

inse

ctio

nse

c.4.

3.B

elow

isan

exam

ple

ofa

topo

logy

whi

chch

ange

sfr

om20

0pr

opan

ols

to20

0pe

ntan

esus

ing

the

GR

OM

OS

-96

forc

efie

ld.

;In

clu

de

forc

efie

ldp

ara

me

ters

#in

clu

de

"ffG

43

a1

.itp

"

[m

ole

cule

typ

e]

;N

am

en

rexc

lP

rop

Pe

nt

3

[a

tom

s]

;n

rty

pe

resn

rre

sid

ue

ato

mcg

nr

cha

rge

ma

ssty

pe

Bch

arg

eB

ma

ssB

1H

1P

RO

PP

H1

0.3

98

1.0

08

CH

30

.01

5.0

35

2O

A1

PR

OP

PO

1-0

.54

81

5.9

99

4C

H2

0.0

14

.02

73

CH

21

PR

OP

PC

11

0.1

50

14

.02

7C

H2

0.0

14

.02

74

CH

21

PR

OP

PC

22

0.0

00

14

.02

75

CH

31

PR

OP

PC

32

0.0

00

15

.03

5

[b

on

ds

];

ai

aj

fun

ctp

ar_

Ap

ar_

B1

22

gb

_1

gb

_2

62

32

gb

_1

7g

b_

26

34

2g

b_

26

gb

_2

64

52

gb

_2

6

[p

airs

];

ai

aj

fun

ct1

41

25

1

Page 120: CS USER AL Simulations - GROMACS

10

0C

ha

pte

r5

.To

po

logie

s

[a

ng

les

];

ai

aj

ak

fun

ctp

ar_

Ap

ar_

B1

23

2g

a_

11

ga

_1

42

34

2g

a_

14

ga

_1

43

45

2g

a_

14

ga

_1

4

[d

ihe

dra

ls]

;a

ia

ja

ka

lfu

nct

pa

r_A

pa

r_B

12

34

1g

d_

12

gd

_1

72

34

51

gd

_1

7g

d_

17

[syste

m]

;N

am

eP

rop

an

ol

toP

en

tan

e

[m

ole

cule

s]

;C

om

po

un

d#

mo

lsP

rop

Pe

nt

20

0

Atom

sthat

arenot

perturbed,PC2

andP

C3,

donot

needB

parameter

specifications,the

Bpa-

rameters

willbe

copiedfrom

theA

parameters.

Bonded

interactionsbetw

eenatom

sthat

arenot

perturbeddo

notneed

Bparam

eterspecifications,

herethis

isthe

casefor

thelast

bond.Topolo-

giesusing

theG

RO

MA

CS

forcefield

needno

bondedparam

etersat

all,since

boththe

Aand

Bparam

etersare

determined

bythe

atomtypes.

Non-bonded

interactionsinvolving

oneor

two

per-turbed

atoms

usethe

free-energyperturbation

functionalforms.

Non-bonded

interactionbetw

eentw

onon-perturbed

atoms

usethe

normalfunctionalform

s.T

hism

eansthat

when,

forinstance,

onlythe

chargeofa

particleis

perturbed,itsLennard-Jones

interactionsw

illalsobe

affectedw

henlam

bdais

notequaltozero

orone.

Note

thatthistopology

usesthe

GR

OM

OS

-96force

field,inw

hichthe

bondedinteractions

arenot

determined

bythe

atomtypes.

The

bondedinteraction

stringsare

convertedby

theC

-preprocessor.T

heforce

fieldparam

eterfiles

containlines

like:

#d

efin

eg

b_

26

0.1

53

07

.15

00

e+

06

#d

efin

eg

d_

17

0.0

00

5.8

63

5.6.5C

onstraintforce

The

constraintforce

between

two

atoms

inone

molecule

canbe

calculatedw

iththe

freeenergy

perturbationcode

byadding

aconstraint

between

thetw

oatom

s,w

itha

differentlength

inthe

Aand

Btopology.

When

theB

lengthis

1nanom

eterlonger

thanthe

Alength

andlam

bdais

keptconstant

atzero,

thederivative

ofthe

Ham

iltonianw

ithrespect

tolam

bdais

theconstraint

force.F

orconstraintsbetw

eenm

oleculesthe

pullcodecan

beused,see

sec.6.1.

Below

isan

example

forcalculating

theconstraint

forceat

0.7nanom

eterbetw

eentw

om

ethanesin

water,

bycom

biningthe

two

methanes

intoone

molecule.

The

addedconstraintis

offunctiontype

2,which

means

thatitis

notusedfor

generatingexclusions

(see5.3.5).

;In

clud

efo

rcefie

ldp

ara

me

ters

7.4

.P

rogra

ms

by

top

ic1

29

freezegrps:G

roupsthat

areto

befrozen

(i.e.their

X,

Y,and/or

Zposition

will

notbe

updated;e.g.

Lip

idS

OL

).freezedim

specifiesfor

which

dimension

thefreezing

applies.You

might

wantto

useenergy

groupexclusions

forcom

pletelyfrozen

groups.

freezedim:

dimensions

forw

hichgroups

infreezegrpsshouldbe

frozen,specifyY

orN

forX

,Y

andZ

andfor

eachgroup

(e.g.YY

NN

NNm

eansthatparticles

inthe

firstgroupcan

move

onlyin

Zdirection.

The

particlesin

thesecond

groupcan

move

inany

direction).

cosacceleration:

(0)[nm

ps −2]

theam

plitudeofthe

accelerationprofile

forcalculating

theviscosity.

The

accelerationis

inthe

X-direction

andthe

magnitude

iscosaccelerationcos(2

piz/boxheight).Tw

oterm

sare

addedto

theenergy

file:the

amplitude

ofthevelocity

profileand

1/viscosity.

7.3.19E

lectricfields

Ex

;Ey

;Ez:

Ifyouw

anttouse

anelectric

fieldin

adirection,enter

3num

bersafter

theappropriate

E*,

thefirstnum

ber:the

number

ofcosines,only1

isim

plemented

(with

frequency0)

soenter

1,the

secondnum

ber:the

strengthof

theelectric

fieldin

Vnm

−1,

thethird

number:

thephase

ofthecosine,you

canenter

anynum

berhere

sincea

cosineoffrequency

zerohas

nophase.

Ext;E

yt;Ezt:

notimplem

entedyet

7.3.20U

serdefined

thingies

user1grps

;user2grps:

userint1(0);userint2

(0);userint3(0);userint4:

(0)

userreal1(0);userreal2

(0);userreal3(0);userreal4:

(0)T

heseyou

canuse

ifyouhack

outcode.You

canpass

integersand

realsto

yoursubroutine.

Check

theinputrec

definitioninsrc/in

clud

e/typ

es/in

pu

trec.h

7.4P

rograms

bytopic

Generating

topologiesand

coordinatespdb2gm

xconverts

pdbfiles

totopology

andcoordinate

filesx2top

generatesa

primitive

topologyfrom

coordinateseditconf

editsthe

boxand

writes

subgroupsgenbox

solvatesa

system

Page 121: CS USER AL Simulations - GROMACS

12

8C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

disr

efc

:(1

000)

[kJ

mol−

1nm

−2]

forc

eco

nsta

ntfo

rdi

stan

cere

stra

ints

,whi

chis

mul

tiplie

dby

a(p

ossi

bly)

diffe

rent

fact

orfo

rea

chre

stra

int

disr

eta

u:(0

)[p

s]tim

eco

nsta

ntfo

rdi

stan

cere

stra

ints

runn

ing

aver

age

nstd

isre

out:

(100

)[s

teps

]fr

eque

ncy

tow

rite

the

runn

ing

time

aver

aged

and

inst

anta

neou

sdi

stan

ces

ofal

lato

mpa

irsin

volv

edin

rest

rain

tsto

the

ener

gyfil

e(c

anm

ake

the

ener

gyfil

eve

ryla

rge)

7.3.

17F

ree

Ene

rgy

Per

turb

atio

n

free

ener

gy:

noO

nly

use

topo

logy

A.

yes

Inte

rpol

ate

betw

een

topo

logy

A(la

mbd

a=0)

toto

polo

gyB

(lam

bda=

1)an

dw

rite

the

deriv

ativ

eof

the

Ham

ilton

ian

with

resp

ectt

ola

mbd

ato

the

ener

gyfil

ean

dto

dg

dl.x

vg.

The

pote

ntia

ls,

bond

-leng

ths

and

angl

esar

ein

terp

olat

edlin

early

asde

scrib

edin

the

man

ual.

Whe

nsc

alph

ais

larg

erth

anze

ro,

soft-

core

pote

ntia

lsar

eus

edfo

rth

eLJ

and

Cou

lom

bin

tera

ctio

ns.

init

lam

bda:

(0)

star

ting

valu

efo

rla

mbd

a

delta

lam

bda:

(0)

incr

ease

per

time

step

for

lam

bda

scal

pha:

(0)

the

soft-

core

para

met

er,

ava

lue

of0

resu

ltsin

linea

rin

terp

olat

ion

ofth

eLJ

and

Cou

lom

bin

tera

ctio

ns

scsi

gma:

(0.3

)[n

m]

the

soft-

core

sigm

afo

rpa

rtic

les

whi

chha

vea

C6

orC

12pa

ram

eter

equa

lto

zero

7.3.

18N

on-e

quili

briu

mM

D

acc

grps

:gr

oups

for

cons

tant

acce

lera

tion

(e.g

.:P

rote

inS

ol

)al

lato

ms

ingr

oups

Pro

tein

and

Sol

will

expe

rienc

eco

nsta

ntac

cele

ratio

nas

spec

ified

inth

eac

cele

rate

line

acce

lera

te:

(0)

[nm

ps−

2]

acce

lera

tion

fora

ccgr

ps;

x,y

and

zfo

rea

chgr

oup

(e.g

.0.1

0.0

0.0

-0.1

0.0

0.0

mea

nsth

atfir

stgr

oup

has

cons

tant

acce

lera

tion

of0.

1nm

ps−

2in

Xdi

rect

ion,

seco

ndgr

oup

the

oppo

site

).

5.6

.F

ilefo

rma

ts1

01

#in

clu

de

"ffG

43

a1

.itp

"

[m

ole

cule

typ

e]

;N

am

en

rexc

lM

eth

an

es

1

[a

tom

s]

;n

rty

pe

resn

rre

sid

ua

tom

cgn

rch

arg

em

ass

1C

H4

1C

H4

C1

10

16

.04

32

CH

41

CH

4C

22

01

6.0

43

[co

nst

rain

ts]

;a

ia

jfu

nct

leng

th_

Ale

ng

th_

B1

22

0.7

1.7

#in

clu

de

"sp

c.itp

"

[sy

ste

m]

;N

am

eM

eth

an

es

inW

ate

r

[m

ole

cule

s]

;C

om

po

un

d#

mo

lsM

eth

an

es

1S

OL

20

02

5.6.

6C

oord

inat

efil

e

File

sw

ithth

e.g

rofil

eex

tens

ion

cont

ain

am

olec

ular

stru

ctur

ein

GR

OM

OS

87fo

rmat

.A

sam

ple

piec

eis

incl

uded

belo

w:

MD

of

2w

ate

rs,

refo

rma

tst

ep

,P

Aa

ug

-91

6 1W

AT

ER

OW

11

0.1

26

1.6

24

1.6

79

0.1

22

7-0

.05

80

0.0

43

41

WA

TE

RH

W2

20

.19

01

.66

11

.74

70

.80

85

0.3

19

1-0

.77

91

1W

AT

ER

HW

33

0.1

77

1.5

68

1.6

13

-0.9

04

5-2

.64

69

1.3

18

02

WA

TE

RO

W1

41

.27

50

.05

30

.62

20

.25

19

0.3

14

0-0

.17

34

2W

AT

ER

HW

25

1.3

37

0.0

02

0.6

80

-1.0

64

1-1

.13

49

0.0

25

72

WA

TE

RH

W3

61

.32

60

.12

00

.56

81

.94

27

-0.8

21

6-0

.02

44

1.8

20

60

1.8

20

60

1.8

20

60

Thi

sfo

rmat

isfix

ed,i

.e.a

llco

lum

nsar

ein

afix

edpo

sitio

n.If

you

wan

tto

read

such

afil

ein

your

own

prog

ram

with

outu

sing

the

GR

OM

AC

Slib

rarie

syo

uca

nus

eth

efo

llow

ing

form

ats:

C-f

orm

at:

"%5

i%5

s%5

s%5

i%8

.3f%

8.3

f%8

.3f%

8.4

f%8

.4f%

8.4

f"

Or

tobe

mor

epr

ecis

e,w

ithtit

leet

c.,i

tloo

kslik

eth

is:

"%s\

n",

Titl

e"%

5d

\n",

na

tom

sfo

r(i=

0;

(i<

na

tom

s);

i++

){

Page 122: CS USER AL Simulations - GROMACS

10

2C

ha

pte

r5

.To

po

logie

s

"%5

d%

5s%

5s%

5d

%8

.3f%

8.3

f%8

.3f%

8.4

f%8

.4f%

8.4

f\n",

resid

ue

nr,re

sidu

en

am

e,a

tom

na

me

,ato

mn

r,x,y,z,vx,vy,vz}"%

10

.5f%

10

.5f%

10

.5f%

10

.5f%

10

.5f%

10

.5f%

10

.5f%

10

.5f%

10

.5f\n

",b

ox[X

][X],b

ox[Y

][Y],b

ox[Z

][Z],

bo

x[X][Y

],bo

x[X][Z

],bo

x[Y][X

],bo

x[Y][Z

],bo

x[Z][X

],bo

x[Z][Y

]

Fortran

format:

(i5,2

a5

,i5,3

f8.3

,3f8

.4)

So

con

fin.g

rois

theG

RO

MA

CS

coordinatefile

andis

almost

thesam

eas

theG

RO

MO

S-

87file

(forG

RO

MO

Susers:

when

usedw

ithntx=

7).T

heonly

differenceis

thebox

forw

hichG

RO

MA

CS

usesa

tensor,notavector.

7.3

.R

un

Pa

ram

ete

rs1

27

morse:no

bondsare

representedby

aharm

onicpotential

yesbonds

arerepresented

bya

Morse

potential

7.3.15E

nergygroup

exclusions

energygrpexcl:

Pairs

ofenergygroups

forw

hichallnon-bonded

interactionsare

excluded.A

nexam

ple:if

youhave

two

energygroupsPro

tein

andS

OL,specifying

en

erg

ye

xcl=

Pro

tein

Pro

tein

SO

LS

OL

would

giveonly

thenon-bonded

interactionsbetw

eenthe

proteinand

thesolvent.

This

isespecially

usefulforspeedingup

energycalculations

with

md

run

-reru

nand

forexclud-ing

interactionsw

ithinfrozen

groups.

7.3.16N

MR

refinement

disre:nono

distancerestraints

(ignoredistance

restraintsinform

ationin

topologyfile)

simplesim

ple(per-m

olecule)distance

restraints

ensemble

distancerestraints

overan

ensemble

ofmolecules

disrew

eighting:

equaldividethe

restraintforceequally

overallatom

pairsin

therestraint

conservativethe

forcesare

thederivative

ofthe

restraintpotential,

thisresults

inan

r−

7w

eightingofthe

atompairs

disrem

ixed:

nothe

violationused

inthe

calculationofthe

restraintforceis

thetim

eaveraged

violation

yesthe

violationused

inthe

calculationofthe

restraintforceis

thesquare

rootofthetim

eaveraged

violationtim

esthe

instantaneousviolation

Page 123: CS USER AL Simulations - GROMACS

12

6C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

7.3.

14B

onds

cons

trai

nts:

none

No

cons

trai

nts,

i.e.b

onds

are

repr

esen

ted

bya

harm

onic

ora

Mor

sepo

tent

ial(

depe

nd-

ing

onth

ese

tting

ofmor

se)

and

angl

esby

aha

rmon

icpo

tent

ial.

hbon

ds Onl

yco

nstr

ain

the

bond

sw

ithH

-ato

ms.

all-b

onds

Con

stra

inal

lbon

ds.

h-an

gles

Con

stra

inal

lbo

nds

and

cons

trai

nth

ean

gles

that

invo

lve

H-a

tom

sby

addi

ngbo

nd-

cons

trai

nts.

all-a

ngle

sC

onst

rain

allb

onds

and

cons

trai

nal

lang

les

byad

ding

bond

-con

stra

ints

.

cons

trai

ntal

gorit

hm:

lincs

LIN

earC

onst

rain

tSol

ver.

The

accu

racy

inse

twithlin

csor

der,

whi

chse

tsth

enu

mbe

rof

mat

rices

inth

eex

pans

ion

for

the

mat

rixin

vers

ion,

4is

enou

ghfo

ra

”nor

mal

”M

Dsi

mul

atio

n,8

isne

eded

forB

Dw

ithla

rge

time-

step

s.T

heac

cura

cyof

the

cons

trai

nts

ispr

inte

dto

the

log

file

ever

ynstlo

gst

eps.

Ifa

bond

rota

tes

mor

eth

anlincs

war

nang

le[d

egre

es]

inon

est

ep,

aw

arni

ngw

illbe

prin

ted

both

toth

elo

gfil

ean

dto

std

err

.Li

ncs

shou

ldno

tbe

used

with

coup

led

angl

eco

nstr

aint

s.

shak

e Sha

keis

slow

eran

dle

ssst

able

than

Linc

s,bu

tdoe

sw

ork

with

angl

eco

nstr

aint

s.T

here

lativ

eto

lera

nce

isse

twiths

hake

tol,

0.00

01is

ago

odva

lue

for

”nor

mal

”M

D.

unco

nstr

aine

dst

art:

noap

ply

cons

trai

nts

toth

est

artc

onfig

urat

ion

yes

dono

tapp

lyco

nstr

aint

sto

the

star

tcon

figur

atio

n

shak

eto

l:(0

.000

1)re

lativ

eto

lera

nce

for

shak

e

lincs

orde

r:(4

)H

ighe

stor

der

inth

eex

pans

ion

ofth

eco

nstr

aint

coup

ling

mat

rix.

lincs

orde

ris

also

used

for

the

num

ber

ofLi

ncs

itera

tions

durin

gen

ergy

min

imiz

atio

n,on

lyon

eite

ratio

nis

used

inM

D.

lincs

war

nang

le:

(30)

[deg

rees

]m

axim

uman

gle

that

abo

ndca

nro

tate

befo

reLi

ncs

will

com

plai

n

Cha

pter

6

Spe

cial

Topi

cs

6.1

Cal

cula

ting

pote

ntia

lsof

mea

nfo

rce:

the

pull

code

The

rear

ea

num

ber

ofop

tions

toca

lcul

ate

pote

ntia

lsof

mea

nfo

rce

and

rela

ted

topi

cs.

Inth

ecu

rren

tver

sion

ofG

RO

MA

CS

this

isim

plem

ente

dth

roug

hso

me

extr

afil

esfo

rm

dru

n.

6.1.

1O

verv

iew

Fou

rdi

ffere

ntty

pes

ofca

lcul

atio

nar

esu

ppor

ted:

1.C

onst

rain

tfor

ces

The

dist

ance

betw

een

the

cent

ers

ofm

ass

oftw

ogr

oups

ofat

oms

can

beco

nstr

aine

dan

dth

eco

nstr

aint

forc

em

onito

red.

The

dist

ance

can

bein

1,2,

or3

dim

ensi

ons.

Thi

sm

etho

dus

esth

eS

HA

KE

algo

rithm

but

only

need

s1

itera

tion

tobe

exac

tif

only

two

grou

psar

eco

nstr

aine

d.

2.U

mbr

ella

sam

plin

gA

sim

ple

umbr

ella

sam

plin

gw

ithan

harm

onic

umbr

ella

pote

ntia

ltha

tac

tson

the

cent

erof

mas

sof

agr

oup

ofat

oms.

3.A

FM

pulli

ngA

sprin

gis

conn

ecte

dto

anat

oman

dsl

owly

retr

acte

d.T

his

has

the

effe

ctof

pulli

ngan

atom

orgr

oup

ofat

oms

away

from

itsin

itial

loca

tion.

The

rate

cons

tant

and

sprin

gco

nsta

ntfo

rth

esp

ring

can

beva

ried

tost

udy

e.g.

the

unbi

ndin

gof

apr

otei

nan

da

ligan

d(s

eefig

ure6

.1).

4.S

tart

ing

stru

ctur

esT

his

optio

ncr

eate

sa

num

bero

fsta

rtin

gst

ruct

ures

forp

oten

tialo

fmea

nfo

rce

calc

ulat

ions

,mov

ing

1or

2gr

oups

ofat

oms

ata

spec

ified

rate

tow

ards

oraw

ayfr

oma

refe

renc

egr

oup,

writ

ing

out

aco

ordi

nate

file

atsp

ecifi

edin

terv

als.

Not

eth

atth

egr

oups

give

nin

the

inde

xfil

ear

etr

ansl

ated

asp

ecifi

eddi

stan

ceea

chst

ep,

but

inad

ditio

nal

soun

derg

oth

eno

rmal

MD

,su

bjec

tto

defin

ition

sofe.g.

tem

pera

ture

coup

ling

grou

ps,

free

zegr

oups

and

the

like.

Inth

eca

lcul

atio

ns,

ther

eha

sto

be1

refe

renc

egr

oup

and

1or

2ot

her

grou

psof

atom

s.F

orco

nstr

aine

dru

ns,

the

dist

ance

betw

een

the

refe

renc

egr

oup

and

the

othe

rgr

oups

iske

ptco

nsta

nt

Page 124: CS USER AL Simulations - GROMACS

10

4C

ha

pte

r6

.S

pe

cialTo

pics

V

zz

linkspring rup

Figure

6.1:S

chematic

pictureof

pullinga

lipidout

ofa

lipidbilayer

with

AF

Mpulling.

Vrup

isthe

velocityatw

hichthe

springis

retracted,Z

link

isthe

atomto

which

thespring

isattached

andZ

sprin

gis

thelocation

ofthespring.

atthedistance

theyhave

inthe

inputcoordinatefile

(.tp

r)

file.

6.1.2U

sage

Inputfiles

The

md

run

programs

needs4

additionalfiles:2

inputfilesand

2outputfiles.

-pi

pu

ll.pp

aIfthis

fileis

specifiedthe

pullcodew

illbeused.

Itcontainsthe

parameters

thatcontrolwhat

typeofcalculation

isdone.

Afullexplanation

ofalltheoptions

isgiven

below.

-pn

ind

ex.n

dx

This

filedefines

thedifferentgroups

foruse

inallpullcalculations.

The

groupsare

referredto

bynam

e,sothe

indexfile

cancontain

othergroups

thatarenotused

asw

ell.

-po

pu

llou

t.pp

aA

formatted

copyofthe

inputparameter

filew

iththe

parameters

thatwere

actuallyused

inthe

run.

-pd

op

ull.p

do

The

datafile

with

thecalculated

forces(A

FM

pulling,constraint

force)or

positions(um

-brella

sampling).

Definition

ofgroups

The

way

thereference

groupsand

differentreference

typesw

orkis

summ

arizedin

figure6.2.

There

arefour

differentpossibilitiesfor

thereference

group.

7.3

.R

un

Pa

ram

ete

rs1

25

compressibility:

[bar −1

]com

pressibility(N

OT

E:

thisis

nowreally

inbar−

1)F

orw

aterat

1atm

and300

Kthe

compressibility

is4.5e-5

[bar −1].

refp:

[bar]

referencepressure

forcoupling

7.3.12S

imulated

annealing

annealing:

noN

osim

ulatedannealing.

yesS

imulated

annealingto

0[K

]at

timezero

temp

time

(ps).R

eferencetem

peraturefor

theB

erendsen-thermostat

isreft

x(1

-tim

e/zero

temp

time),

time

constantis

taut

[ps].N

otethat

thereference

temperature

willnot

gobelow

0[K

],i.e.

afterzero

temp

time

(ifit

ispositive)

thereference

temperature

will

be0

[K].

Negative

zerotem

ptim

eresults

inheating,w

hichw

illgoon

indefinitely.

zerotem

ptim

e:(0)

[ps]tim

eatw

hichtem

peraturew

illbezero

(canbe

negative).Tem

peratureduring

therun

canbe

seenas

astraightline

goingthrough

T=

reft[K

]att=0

[ps],andT

=0

[K]att=zero

temp

time

[ps].Look

inour

FAQ

fora

schematic

graphoftem

peratureversus

time.

7.3.13V

elocitygeneration

genvel:

noD

onotgenerate

velocitiesatstartup.

The

velocitiesare

settozero

when

thereare

novelocities

inthe

inputstructurefile.

yesG

eneratevelocities

accordingto

aM

axwelldistribution

attemperature

gentem

p[K

],w

ithrandom

seedgenseed.

This

isonly

meaningfulw

ithintegratormd.

gentem

p:(300)

[K]

temperature

forM

axwelldistribution

genseed:

(173529)[integer]

usedto

initializerandom

generatorfor

randomvelocities,

whengen

seedisset

to-1,

theseed

iscalculated

as(time

()+

ge

tpid

())%

10

00

00

0

Page 125: CS USER AL Simulations - GROMACS

12

4C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

noN

opr

essu

reco

uplin

g.T

his

mea

nsa

fixed

box

size

.

bere

ndse

nE

xpon

entia

lrel

axat

ion

pres

sure

coup

ling

with

time

cons

tant

tau

p[p

s].

The

box

issc

aled

ever

ytim

este

p.It

has

been

argu

edth

atth

isdo

esno

tyie

lda

corr

ectt

herm

ody-

nam

icen

sem

ble,

but

itis

the

mos

tef

ficie

ntw

ayto

scal

ea

box

atth

ebe

ginn

ing

ofa

run.

Par

inel

lo-R

ahm

anE

xten

ded-

ense

mbl

epr

essu

reco

uplin

gw

here

the

box

vect

ors

are

subj

ectt

oan

equa

tion

ofm

otio

n.T

heeq

uatio

nof

mot

ion

for

the

atom

sis

coup

led

toth

is.

No

inst

anta

neou

ssc

alin

gta

kes

plac

e.A

sfo

rNos

e-H

oove

rtem

pera

ture

coup

ling

the

time

cons

tant

tau

p[p

s]is

the

perio

dof

pres

sure

fluct

uatio

nsat

equi

libriu

m.

Thi

sis

prob

ably

abe

tter

met

hod

whe

nyo

uw

ant

toap

ply

pres

sure

scal

ing

durin

gda

taco

llect

ion,

but

bew

are

that

you

can

getv

ery

larg

eos

cilla

tions

ifyo

uar

est

artin

gfr

oma

diffe

rent

pres

sure

.

pcou

plty

pe:

isot

ropi

cIs

otro

pic

pres

sure

coup

ling

with

time

cons

tanttau

p[p

s].

The

com

pres

sibi

lity

and

refe

renc

epr

essu

rear

ese

tw

ithcom

pres

sibi

lity

[bar−

1]

and

ref

p[b

ar],

one

valu

eis

need

ed.

sem

iisot

ropi

cP

ress

ure

coup

ling

whi

chis

isot

ropi

cin

the

xan

dy

dire

ctio

n,bu

tdi

ffere

ntin

the

zdi

rect

ion.

Thi

sca

nbe

usef

ulfo

rm

embr

ane

sim

ulat

ions

.2

valu

esar

ene

eded

for

x/y

and

zdi

rect

ions

resp

ectiv

ely.

anis

otro

pic

Idem

,bu

t6

valu

esar

ene

eded

for

xx,

yy,

zz,

xy/y

x,xz

/zx

and

yz/z

yco

mpo

nent

sre

spec

tivel

y.W

hen

the

off-

diag

onal

com

pres

sibi

litie

sar

ese

tto

zero

,are

ctan

gula

rbox

will

stay

rect

angu

lar.

Bew

are

that

anis

otro

pic

scal

ing

can

lead

toex

trem

ede

form

atio

nof

the

sim

ulat

ion

box.

surf

ace-

tens

ion

Sur

face

tens

ion

coup

ling

for

surf

aces

para

llelt

oth

exy

-pla

ne.

Use

sno

rmal

pres

sure

coup

ling

for

the

z-di

rect

ion,

whi

leth

esu

rfac

ete

nsio

nis

coup

led

toth

ex/

ydi

men

-si

ons

ofth

ebo

x.T

hefir

stref

pva

lue

isth

ere

fere

nce

surf

ace

tens

ion

times

the

num

-be

rof

surf

aces

[bar

nm],

the

seco

ndva

lue

isth

ere

fere

nce

z-pr

essu

re[b

ar].

The

two

com

pres

sibi

lity

[bar−

1]

valu

esar

eth

eco

mpr

essi

bilit

yin

the

x/y

and

zdi

rect

ion

re-

spec

tivel

y.T

heva

lue

for

the

z-co

mpr

essi

bilit

ysh

ould

bere

ason

ably

accu

rate

sinc

eit

influ

ence

sth

eco

nver

geof

the

surf

ace-

tens

ion,

itca

nal

sobe

sett

oze

roto

have

abo

xw

ithco

nsta

nthe

ight

.

tric

linic Ful

lydy

nam

icbo

x-

supp

orte

d,bu

tex

perim

enta

l.Yo

ush

ould

prov

ide

six

valu

esfo

rth

eco

mpr

essi

bilit

yan

dre

fere

nce

pres

sure

.

tau

p:(1

)[p

s]tim

eco

nsta

ntfo

rco

uplin

g

6.1

.C

alc

ula

ting

po

ten

tials

ofm

ea

nfo

rce

:th

ep

ull

cod

e1

05

��

��

Fig

ure

6.2:

Ove

rvie

wof

the

diffe

rent

refe

renc

egr

oup

poss

ibili

ties,

appl

ied

toin

terf

ace

syst

ems.

Cis

the

refe

renc

egr

oup.

The

circ

les

repr

esen

tth

ece

nter

ofm

ass

of2

grou

pspl

usth

ere

fere

nce

grou

p,an

ddc

isth

ere

fere

nce

dist

ance

.

com

The

cent

erof

mas

sof

the

grou

pgi

ven

unde

rre

fere

nce

gro

up

,ca

lcul

ated

each

step

from

the

curr

entc

oord

inat

es.

com

t0 The

cent

erof

mas

sof

the

grou

pgi

ven

unde

rre

fere

nce

gro

up

,ca

lcul

ated

each

step

from

the

curr

ent

coor

dina

tes,

but

corr

ecte

dfo

rat

oms

that

have

cros

sed

the

box.

Ifth

ere

f-er

ence

grou

pco

nsis

tsof

allt

hew

ater

mol

ecul

esin

the

syst

em,a

nda

sing

lew

ater

mol

ecul

em

oves

acro

ssth

ebo

xan

den

ters

from

the

othe

rsi

de,

the

c.o.

m.

will

show

asl

ight

jum

p.T

his

issi

mpl

ydu

eto

the

perio

dic

boun

dary

cond

ition

s,an

dsh

ows

that

the

cent

erof

mas

sin

asi

mul

atio

nin

perio

dic

boun

dary

cond

ition

sis

illde

fined

ifth

egr

oup

used

toca

lcul

ate

itis

e.g.

asl

abof

liqui

d.If

the

’real

’pos

ition

sar

eus

edin

stea

dof

the

coor

dina

tes

that

have

been

rese

tto

bein

side

the

box,

the

cent

erof

mas

sof

the

who

lesy

stem

isco

nser

ved.

dyn

am

ic Ina

phos

phol

ipid

bila

yer

syst

emit

may

beof

inte

rest

toca

lcul

ate

the

pmf

ofa

lipid

asfu

nctio

nof

itsdi

stan

cefr

omth

ew

hole

bila

yer.

The

who

lebi

laye

rca

nbe

take

nas

refe

renc

egr

oup

inth

atca

se,

but

itm

ight

also

beof

inte

rest

tode

fine

the

reac

tion

coor

dina

tefo

rth

epm

fm

ore

loca

lly.d

yna

mic

does

not

use

allt

heat

oms

ofth

erefe

ren

ceg

rou

p,

but

inst

ead

only

thos

ew

ithin

acy

linde

rw

ithra

diusr

belo

wth

em

ain

grou

p.T

his

only

wor

ksfo

rdi

stan

ces

defin

edin

1di

men

sion

,an

dth

ecy

linde

ris

orie

nted

with

itslo

ngax

isal

ong

this

1di

men

sion

.A

seco

ndcy

linde

rca

nbe

defin

edw

ithrc

,w

itha

linea

rsw

itch

func

tion

that

wei

ghs

the

cont

ribut

ion

ofat

oms

betw

een

ran

drc

with

dist

ance

.T

his

smoo

thes

the

effe

cts

ofat

oms

mov

ing

inan

dou

tof

the

cylin

der

(whi

chca

uses

jum

psin

the

cons

trai

ntfo

rces

).

dyn

am

ict0

The

sam

easd

yna

mic

,but

the

coor

dina

tes

are

corr

ecte

dfo

rbo

xcro

ssin

gslik

ein

com

t0.

Page 126: CS USER AL Simulations - GROMACS

10

6C

ha

pte

r6

.S

pe

cialTo

pics

Note

thatstrictlyspeaking

thisis

notcorrectifthereference

groupis

notthew

holesystem

,including

thegroups

definedw

ithgrou

p1

andg

rou

p2

.

Tofurther

smooth

rapidlyfluctuating

distancesbetw

eenthe

referencegroup

andthe

othergroups,

theaverage

distancecan

beconstrained

insteadof

theinstanteneous

distance.T

hisis

definedby

settingre

flag

tothe

number

ofstepsto

averageover.

How

ever,usingthis

optionis

notstrictlycorrectfor

calculatingpotentials

ofmean

forcefrom

theaverage

constraintforce.

The

parameter

file

verb

ose

=n

oIf

thisis

setto

yes

,a

largeam

ountof

detailedinform

ationis

sentto

stde

rr,

which

isonly

usefulfordiagnostic

purposes.T

he.p

do

filealso

becomes

more

detailed,w

hichis

notnecessaryfor

normaluse.

run

type

=co

nstra

int

Options

arestart,

afm

,co

nstra

int,

um

bre

lla.

This

selectsthe

typeof

cal-culation:

making

startingstructures,A

FM

pulling,constraintforcecalculation

orum

brellasam

pling.

gro

up

1=

MB

21

1

gro

up

2=

MB

21

2T

hegroups

with

theatom

sto

acton.T

hefirstgroup

ism

andatory,thesecond

optional.

refe

ren

ceg

rou

p=

OC

TA

The

referencegroup.

Distances

arecalculated

betweeeng

rou

p1

(andg

rou

p2

ifspec-

ified)and

thisgroup.

Ife.g.the

constraintforce

between

two

ionsis

needed,you

would

specifiygro

up

1as

agroup

with

1ion,andre

fere

nce

gro

up

asthe

otherion.

reftyp

e=

com

The

typeofreference

group.O

ptionsareco

m,

com

t0,

dyn

am

ic,d

yna

mic

t0as

explainedabove.

refla

g=

1T

heposition

ofthereference

groupcan

betaken

asaverage

overanum

berofsteps,specifiedby

refla

g(see

above).

dire

ction

=0

.00

.01

.0D

istancesare

calculatedw

eightedby

x,y,

zas

specifiedin

dire

ction

.S

ettingthem

allto

1.0calculates

thedistance

between

two

groups,setting

thefirst

two

to0.0

andthe

thirdto

1.0calculates

thedistance

inthe

zdirection

only.

reve

rse=

tore

fere

nce

This

optionselects

thedirection

inw

hichthe

groupsare

moved

with

respecttothe

referencegroup

forA

FM

pullingand

startingstructure

calculations.C

hoicesare

tore

fere

nce

,fro

mre

fere

nce

.

7.3

.R

un

Pa

ram

ete

rs1

23

pme

order(4)

Interpolationorder

forP

ME

.4equals

cubicinterpolation.

Youm

ighttry6/8/10

when

run-ning

inparalleland

simultaneously

decreasegrid

dimension.

ewald

rtol(1e-5)T

herelative

strengthofthe

Ew

ald-shifteddirectpotentialatthe

cutoffisgiven

byew

aldrtol.

Decreasing

thisw

illgivea

more

accuratedirectsum

,butthenyou

needm

orew

avevectors

forthe

reciprocalsum.

optimize

fft:

noD

on’tcalculatethe

optimalF

FT

planfor

thegrid

atstartup.

yesC

alculatethe

optimalF

FT

planfor

thegrid

atstartup.

This

savesa

fewpercent

forlong

simulations,buttakes

acouple

ofminutes

atstart.

7.3.10Tem

peraturecoupling

tcoupl:noN

otem

peraturecoupling.

berendsenTem

peraturecoupling

with

aB

erendsen-thermostat

toa

bathw

ithtem

peratureref

t[K

],with

time

constanttaut

[ps].S

everalgroupscan

becoupled

separately,theseare

specifiedin

thetcgrps

fieldseparated

byspaces.

nose-hooverTem

peraturecoupling

with

aby

usinga

Nose-H

ooverextended

ensemble.

The

refer-ence

temperature

andcoupling

groupsare

selectedas

above,butinthis

casetau

t[ps]

controlsthe

periodof

thetem

peraturefluctuations

atequilibrium

,w

hichis

slightlydifferentfrom

arelaxation

time.

tcgrps:groups

tocouple

separatelyto

temperature

bath

taut:

[ps]

time

constantforcoupling

(onefor

eachgroup

intc

grps)

reft:

[K]

referencetem

peraturefor

coupling(one

foreach

groupin

tcgrps)

7.3.11P

ressurecoupling

pcoupl:

Page 127: CS USER AL Simulations - GROMACS

12

2C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

vdw

type

:

Cut

-off Tw

inra

nge

cut-

off’s

with

neig

hbor

list

cut-

offrli

stan

dV

dWcu

t-of

frvd

w,

whe

rerv

dw>

rlist

.

Shi

ftT

heLJ

(not

Buc

king

ham

)po

tent

iali

sde

crea

sed

over

the

who

lera

nge

and

the

forc

esde

cay

smoo

thly

toze

robe

twee

nrvdw

switc

han

drv

dw.

The

neig

hbor

sear

chcu

t-of

frli

stsh

ould

be0.

1to

0.3

nmla

rger

thanrv

dwto

acco

mm

odat

efo

rth

esi

zeof

char

gegr

oups

and

diffu

sion

betw

een

neig

hbor

listu

pdat

es.

Use

r md

run

will

now

expe

ctto

find

two

files

with

user

-defi

ned

func

tions

:rt

ab

.xvg

for

Rep

ulsi

on,d

tab

.xvg

for

Dis

pers

ion.

The

sefil

essh

ould

cont

ain

5co

lum

ns:

the

xva

lue,

f(x)

,-f

(1) (

x),

f(2

) (x)

and

-f(3

) (x)

,w

here

f(n

) (x)

deno

tes

then

th

deriv

ativ

eof

func

tionf

(x)

with

resp

ect

tox

.T

hex

shou

ldru

nfr

om0

[nm

]to

rlist

+0

.5[n

m],

with

asp

acin

gof

0.0

02

[nm

]w

hen

you

run

insi

ngle

prec

isio

n,or

0.0

00

5[n

m]

whe

nyo

uru

nin

doub

lepr

ecis

ion.

The

func

tion

valu

eat

x=0

isno

tim

port

ant.

Whe

nyo

uw

ant

tous

eLJ

corr

ectio

n,m

ake

sure

that

rvdw

corr

espo

nds

toth

ecu

t-of

fin

the

user

-defi

ned

func

tion.

rvdw

switc

h:(0

)[n

m]

whe

reto

star

tsw

itchi

ngth

eLJ

pote

ntia

l

rvdw

:(1

)[n

m]

dist

ance

for

the

LJor

Buc

king

ham

cut-

off

Dis

pCor

r:

nodo

n’ta

pply

any

corr

ectio

n

Ene

rPre

sap

ply

long

rang

edi

sper

sion

corr

ectio

nsfo

rE

nerg

yan

dP

ress

ure

Ene

r appl

ylo

ngra

nge

disp

ersi

onco

rrec

tions

for

Ene

rgy

only

four

iers

paci

ng:

(0.1

2)[n

m]

The

max

imum

grid

spac

ing

fort

heF

FT

grid

whe

nus

ing

PP

PM

orP

ME

.For

ordi

nary

Ew

ald

the

spac

ing

times

the

box

dim

ensi

ons

dete

rmin

esth

ehi

ghes

tm

agni

tude

tous

ein

each

di-

rect

ion.

Inal

lca

ses

each

dire

ctio

nca

nbe

over

ridde

nby

ente

ring

ano

n-ze

rova

lue

for

four

ier

n*.

four

ier

nx(0

);f

ourie

rny

(0)

;fou

rier

nz:

(0)

Hig

hest

mag

nitu

deof

wav

eve

ctor

sin

reci

proc

alsp

ace

whe

nus

ing

Ew

ald.

Grid

size

whe

nus

ing

PP

PM

orP

ME

.The

seva

lues

over

ride

four

iers

paci

ngpe

rdi

rect

ion.

The

best

choi

ceis

pow

ers

of2,

3,5

and

7.A

void

larg

epr

imes

.

6.1

.C

alc

ula

ting

po

ten

tials

ofm

ea

nfo

rce

:th

ep

ull

cod

e1

07

r=

0 Ifdy

nam

icre

fere

nce

grou

psar

ese

lect

ed(

dyn

am

ic,

dyn

am

ict0

),r

isth

era

dius

ofth

ecy

linde

rus

edto

defin

ew

hich

atom

sar

epa

rtof

the

refe

renc

egr

oup

(see

abov

e).

rc=

0 With

dyna

mic

refe

renc

egr

oups

,th

ecy

linde

rca

nbe

smoo

thly

switc

hed

soth

atat

oms

that

fall

betw

eenr

and

rcar

ew

eigh

ted

linea

rlyfr

om1

to0

goin

gfr

omrto

rc.

As

reas

onab

lein

itial

valu

esw

esu

gges

tr=

1.0

and

rc=

1.5

but

this

will

depe

ndst

rong

lyon

the

exac

tsys

tem

ofin

tere

st.

up

da

te=

1T

hefr

eque

ncy

with

whi

chth

edy

nam

icre

fere

nce

grou

psar

ere

calc

ulat

ed.

Usu

ally

ther

eis

nore

ason

tous

ean

ythi

ngot

her

than

1.

pu

llra

te=

0.0

00

05

The

pull

rate

innm

/tim

este

pfo

rA

FM

pulli

ng.

forc

eco

nst

an

t=

10

0T

hefo

rce

cons

tant

for

the

sprin

gin

AF

Mpu

lling

,in

kJm

ol−

1nm

−2.

wid

th=

0W

idth

ofth

eum

brel

lasa

mpl

ing

pote

ntia

lin

kJm

ol−

1nm

−2.

r0g

rou

p2

=0

.00

.03

.30

0T

hein

itial

loca

tion

ofth

egr

oups

with

resp

ect

toth

ere

fere

nce

grou

p.O

nly

coor

dina

tes

sele

cted

with

dire

ctio

nar

eta

ken

into

acco

unt.

The

grou

psar

em

oved

toth

ese

initi

alpo

sitio

nsbe

fore

the

actu

alcr

eatio

nof

ase

ries

ofst

artin

gst

ruct

ures

com

men

ces.

tole

ran

ce=

0.0

01

The

accu

racy

with

whi

chth

eac

tual

posi

tion

ofth

egr

oups

mus

tm

atch

the

calc

ulat

edid

eal

posi

tions

for

ast

artin

gst

ruct

ure

(innm

).

tra

nsl

atio

nra

te=

0.0

00

01

The

rate

oftr

ansl

atio

nin

alld

irect

ions

(nm

/ste

p).

As

men

tione

dab

ove,

norm

alM

Dfo

rce

calc

ulat

ions

and

posi

tion

upda

tes

also

acto

nth

egr

oups

.

tra

nss

tep

=0

.2T

hein

terv

alin

nmat

whi

chst

ruct

ures

are

writ

ten

out.

6.1.

3O

utpu

t

The

outp

utfil

eis

ate

xtfil

ew

ithfo

rces

orpo

sitio

ns,

one

per

line.

Ifth

ere

are

two

grou

psth

eyal

tern

ate

inth

eou

tput

file.

Cur

rent

lyth

ere

isno

supp

orte

dan

alys

ispr

ogra

mto

read

this

file,

buti

tis

sim

ple

topa

rse.

6.1.

4Li

mita

tions

Apa

rtfr

omob

viou

slim

itatio

nsth

atar

esi

mpl

yno

tim

plem

ente

d(

e.g.

abe

tter

umbr

ella

sam

plin

gan

dan

alys

issc

hem

e),

ther

eis

one

impo

rtan

tlim

itatio

n:co

nstr

aint

forc

esca

non

lybe

calc

ulat

ed

Page 128: CS USER AL Simulations - GROMACS

10

8C

ha

pte

r6

.S

pe

cialTo

pics

between

molecules

orgroups

ofm

olecules.If

agroup

containspart

ofa

molecule

ofw

hichthe

bondlengthsare

constrained,S

HA

KE

orLIN

CS

andthe

constraintforce

calculationhere

will

interferew

itheach

other,m

akingthe

resultsunreliable.

Ifa

constraintforce

isw

antedbetw

eentw

oatom

s,thiscan

bedone

throughthe

freeenergy

perturbationcode.

Insum

mary:

•pullcode:betw

eenm

oleculesor

groupsofm

olecules.

•free

energyperturbation

code:between

singleatom

s.

•not

possiblecurrently:

between

groupsof

atoms

thatare

partof

alarger

molecule

forw

hichthe

bondsare

constrainedw

ithS

HA

KE

orLIN

CS

.

6.1.5Im

plementation

The

codefor

theoptions

describedabove

canbe

foundin

thefiles

pu

ll.c,p

ullin

it.c,p

ullio

.c,p

ullu

til.cand

theheaderfilespu

ll.hand

pu

lls.h.

This

lastfiledefines

afew

datatypes,pull.h

explainsthe

main

functions.

6.1.6F

uturedevelopm

ent

There

areseveraladditionalfeatures

thatwould

beuseful,including

more

advancedum

brellasam

-pling,an

analysistoolto

analysethe

outputofthepullcode,incorporation

oftheinputparam

etersand

indexfile

intotheg

rom

pp

programinputfiles,extension

tom

oregroups,m

oreflexible

defi-nition

ofareaction

coordinate,extensionto

groupsthatare

partsofm

oleculesthatuse

SH

AK

Eor

LINC

S,and

acom

binationofthe

startingstructure

calculationw

ithconstraints

forfaster

conver-gence

ofstartingstructures.

6.2R

emoving

fastestdegreesoffreedom

The

maxim

umtim

estep

inM

Dsim

ulationsis

limited

bythe

smallest

oscillationperiod

thatcan

befound

inthe

simulated

system.

Bond-stretching

vibrationsare

intheir

quantum-m

echanicalground

stateand

aretherefore

betterrepresented

bya

constraintthanby

aharm

onicpotential.

For

therem

ainingdegrees

offreedom

,the

shortestoscillation

periodas

measured

froma

simu-

lationis

13fs

forbond-angle

vibrationsinvolving

hydrogenatom

s.Taking

asa

guidelinethat

with

aVerlet(leap-frog)integration

scheme

am

inimum

of5num

ericalintegrationsteps

shouldbe

performed

perperiod

ofa

harmonic

oscillationin

orderto

integrateit

with

reasonableaccuracy,

them

aximum

time

stepw

illbeabout3

fs.D

isregardingthese

veryfastoscillations

ofperiod13

fsthe

nextshortestperiodsare

around20

fs,which

willallow

am

aximum

time

stepofabout4

fs

Rem

ovingthe

bond-angledegrees

offreedom

fromhydrogen

atoms

canbest

bedone

bydefin-

ingthem

asdum

my

atoms

instead

ofnorm

alatoms.

Where

anorm

alatoms

isconnected

tothe

molecule

with

bonds,angles

anddihedrals,

adum

my

atom’s

positionis

calculatedfrom

thepo-

sitionof

threenearby

heavyatom

sin

apredefined

manner

(seealso

sec.4.5).

For

thehydrogens

inw

aterand

inhydroxyl,sulfhydrylor

amine

groups,nodegrees

offreedomcan

berem

oved,be-cause

rotationalfreedomshould

bepreserved.

The

onlyother

optionavailable

toslow

down

these

7.3

.R

un

Pa

ram

ete

rs1

21

PM

EF

astP

article-Mesh

Ew

aldelectrostatics.

Direct

spaceis

similar

tothe

Ew

aldsum

,w

hilethe

reciprocalpartisperform

edw

ithF

FT

s.G

riddim

ensionsare

controlledw

ithfourierspacing

andthe

interpolationorder

withpm

eorder.

With

agrid

spacingof

0.1nm

andcubic

interpolationthe

electrostaticforces

havean

accuracyof

2-3e-4.S

incethe

errorfrom

thevdw

-cutoffis

largerthan

thisyou

might

try0.15

nm.

When

runningin

paralleltheinterpolation

parallelizesbetter

thanthe

FF

T,sotry

decreasinggrid

dimensions

while

increasinginterpolation.

PP

PMP

article-Particle

Particle-M

eshalgorithm

forlong

rangeelectrostatic

interactions.U

seforexam

plerlist=1

.0,rcoulom

bsw

itch=0

.0,rcoulom

b=

0.8

5,rvdw

switch=

1.0

andrvdw

=1

.0.

The

griddim

ensionsare

controlledbyfourierspacing.

Reasonable

gridspacing

forP

PP

Mis

0.05-0.1nm

.S

eeS

hift

forthe

detailsof

theparticle-

particlepotential.

NO

TE

:thepressure

inincorrectw

henusing

PP

PM

.

Reaction-F

ieldR

eactionfield

with

Coulom

bcut-offrcoulom

b,wherercoulom

b>

rvdw>

rlist.T

hedielectric

constantbeyond

thecut-off

isepsilonr.

The

dielectricconstant

canbe

setto

infinityby

settingepsilonr=

0.

Generalized-R

eaction-Field

Generalized

reactionfield

with

Coulom

bcut-off

rcoulomb,w

herercoulomb>

rvdw>

rlist.T

hedielectric

constantbeyond

thecut-off

isepsilon

r.T

heionic

strengthis

computed

fromthe

number

ofcharged(i.e.w

ithnon

zerocharge)

chargegroups.

The

temperature

forthe

GR

Fpotentialis

setwithref

t[K

].

ShiftT

heC

oulomb

potentialisdecreased

overthew

holerange

andthe

forcesdecay

smoothly

tozero

betweenrcoulom

bsw

itchand

rcoulomb.

The

neighborsearch

cut-offrlistshould

be0.1

to0.3

nmlarger

thanrcoulomb

toaccom

modate

forthe

sizeof

chargegroups

anddiffusion

between

neighborlistupdates.

UserS

pecifyrshort

andrlong

tothe

same

value,md

run

will

nowexpect

tofind

afile

ctab

.xvgw

ithuser-defined

functions.T

hisfile

shouldcontain

5colum

ns:the

xvalue,f(x)

,-f

(1)(x)

,f

(2)(x)

and-f

(3)(x)

,w

heref

(n)(x)

denotesthen

th

derivativeof

functionf(x)w

ithrespect

tox.

The

xshould

runfrom

0[nm

]to

rlist+0

.5[nm

],w

itha

spacingof0

.00

2[nm

]w

henyou

runin

singleprecision,

or0

.00

05

[nm]

when

yourun

indouble

precision.T

hefunction

valueat

x=0

isnot

important.

rcoulomb

switch:

(0)[nm

]w

hereto

startswitching

theC

oulomb

potential

rcoulomb:

(1)[nm

]distance

forthe

Coulom

bcut-off

epsilonr:

(1)dielectric

constant

Page 129: CS USER AL Simulations - GROMACS

12

0C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

xtc

grps

:gr

oup(

s)to

writ

eto

xtc

traj

ecto

ry,d

efau

ltth

ew

hole

syst

emis

writ

ten

(ifns

txtc

outi

sla

rger

than

zero

)

ener

gygr

ps:

grou

p(s)

tow

rite

toen

ergy

file

7.3.

8N

eigh

bor

sear

chin

g

nstli

st:

(10)

[ste

ps]

Fre

quen

cyto

upda

teth

ene

ighb

orlis

t(a

ndth

elo

ng-r

ange

forc

es,

whe

nus

ing

twin

-ran

gecu

t-of

f’s).

Whe

nth

isis

0,th

ene

ighb

orlis

tis

mad

eon

lyon

ce.

nsty

pe:

grid

Mak

ea

grid

inth

ebo

xan

don

lych

eck

atom

sin

neig

hbor

ing

grid

cells

whe

nco

nstr

uct-

ing

ane

wne

ighb

orlis

teve

rynst

lists

teps

.In

larg

esy

stem

sgr

idse

arch

ism

uch

fast

erth

ansi

mpl

ese

arch

.

sim

ple C

heck

ever

yat

omin

the

box

whe

nco

nstr

uctin

ga

new

neig

hbor

liste

very

nstli

stst

eps.

pbc:

xyz

Use

perio

dic

boun

dary

cond

ition

sin

alld

irect

ions

.

noU

seno

perio

dic

boun

dary

cond

ition

s,ig

nore

the

box.

Tosi

mul

ate

with

outc

ut-o

ffs,s

etal

lcut

-offs

to0

andn

stlis

t=0.

rlist

:(1

)[n

m]

cut-

offd

ista

nce

for

the

shor

t-ra

nge

neig

hbor

list

7.3.

9E

lect

rost

atic

san

dV

dW

coul

ombt

ype:

Cut

-off Tw

inra

nge

cut-

off’s

with

neig

hbor

list

cut-

offrli

stan

dC

oulo

mb

cut-

offrc

oulo

mb,

whe

rerli

st<

rvdw

<rc

oulo

mb.

The

diel

ectr

icco

nsta

ntis

setw

itheps

ilon

r.

Ew

ald C

lass

ical

Ew

ald

sum

elec

tros

tatic

s.U

see.

g.rli

st=

0.9,

rvdw

=0.

9,rc

oulo

mb=

0.9.

The

high

est

mag

nitu

deof

wav

eve

ctor

sus

edin

reci

proc

alsp

ace

isco

ntro

lled

byfo

uri-

ersp

acin

g.T

here

lativ

eac

cura

cyof

dire

ct/r

ecip

roca

lspa

ceis

cont

rolle

dby

ewal

drt

ol.

NO

TE

:E

wal

dsc

ales

asO

(N3/2)

and

isth

usex

trem

ely

slow

for

larg

esy

stem

s.It

isin

clud

edm

ainl

yfo

rre

fere

nce

-in

mos

tcas

esP

ME

will

perf

orm

muc

hbe

tter.

6.2

.R

em

ovin

gfa

ste

std

egre

es

off

ree

do

m1

09

D

d

α

d BA

C

���������� ����������

���������� ����������

���������� ����������

�����

�����

�����

�����

��

��

��

��

���������� ����������

� ���

�����

�����

�����

�����

���������� ����������

Fig

ure

6.3:

The

diffe

rent

type

sof

dum

my

atom

cons

truc

tions

used

forh

ydro

gen

atom

s.T

heat

oms

used

inth

eco

nstr

uctio

nof

the

dum

my

atom

(s)a

rede

pict

edas

blac

kci

rcle

s,du

mm

yat

oms

asgr

eyon

es.

Hyd

roge

nsar

esm

alle

rth

anhe

avy

atom

s.A

:fixe

dbo

ndan

gle,

note

that

here

the

hydr

ogen

isno

tadu

mm

yat

om;B:i

nth

epl

ane

ofth

ree

atom

s,w

ithfix

eddi

stan

ce;

C:i

nth

epl

ane

ofth

ree

atom

s,w

ithfix

edan

gle

and

dist

ance

;D

:con

stru

ctio

nfo

ram

ine

grou

ps(

-NH

2or

-NH

+ 3),

see

text

for

deta

ils.

mot

ions

,is

toin

crea

seth

em

ass

ofth

ehy

drog

enat

oms

atth

eex

pens

eof

the

mas

sof

the

conn

ecte

dhe

avy

atom

.T

his

will

incr

ease

the

mom

ent

ofin

ertia

ofth

ew

ater

mol

ecul

esan

dth

ehy

drox

yl,

sulfh

ydry

lor

amin

egr

oups

,w

ithou

taf

fect

ing

the

equi

libriu

mpr

oper

ties

ofth

esy

stem

and

with

-ou

taffe

ctin

gth

edy

nam

ical

prop

ertie

sto

om

uch.

The

seco

nstr

uctio

nsw

illsh

ortly

bede

scrib

edin

sec.

6.2.

1an

dha

vepr

evio

usly

been

desc

ribed

infu

llde

tail

[63

].

Usi

ngbo

thdu

mm

yat

oms

and

mod

ified

mas

ses,

the

next

bottl

enec

kis

likel

yto

befo

rmed

byth

eim

prop

erdi

hedr

als

(whi

char

eus

edto

pres

erve

plan

arity

orch

iralit

yof

mol

ecul

argr

oups

)an

dth

epe

ptid

edi

hedr

als.

The

pept

ide

dihe

dral

cann

otbe

chan

ged

with

out

affe

ctin

gth

eph

ysic

albe

-ha

vior

ofth

epr

otei

n.T

heim

prop

erdi

hedr

als

that

pres

erve

plan

arity

,m

ostly

deal

with

arom

atic

resi

dues

.B

onds

,ang

les

and

dihe

dral

sin

thes

ere

sidu

esca

nal

sobe

repl

aced

with

som

ewha

tela

b-or

ate

dum

my

atom

cons

truc

tions

,as

will

bede

scrib

edin

sec.

6.2.

2[6

4].

All

mod

ifica

tions

desc

ribed

inth

isse

ctio

nca

nbe

perf

orm

edus

ing

the

GR

OM

AC

Sto

polo

gybu

ild-

ing

tool

pd

b2

gm

x.

Sep

arat

eop

tions

exis

tto

incr

ease

hydr

ogen

mas

ses,

dum

mify

allh

ydro

gen

atom

sor

also

dum

mify

alla

rom

atic

resi

dues

.N

ote

that

whe

nal

lhyd

roge

nat

oms

are

dum

mifi

ed,

also

thos

ein

side

the

arom

atic

resi

dues

will

bedu

mm

ified

,i.e

.hyd

roge

nsin

the

arom

atic

resi

dues

are

trea

ted

diffe

rent

lyde

pend

ing

onth

etr

eatm

ento

fthe

arom

atic

resi

dues

.

Par

amet

ers

for

the

dum

my

cons

truc

tions

for

the

hydr

ogen

atom

sar

ein

ferr

edfr

omth

efo

rcefi

eld

para

met

ers

(vis.

bond

leng

ths

and

angl

es)

dire

ctly

bygro

mp

pw

hile

proc

essi

ngth

eto

polo

gyfil

e.T

heco

nstr

uctio

nsfo

rth

ear

omat

icre

sidu

esar

eba

sed

onth

ebo

ndle

ngth

san

dan

gles

for

the

geom

etry

asde

scrib

edin

the

forc

efiel

ds,b

utth

ese

para

met

ers

are

hard

-cod

edin

top

db

2g

mx

due

toth

eco

mpl

exna

ture

ofth

eco

nstr

uctio

nne

eded

for

aw

hole

arom

atic

grou

p.

6.2.

1H

ydro

gen

bond

-ang

levi

brat

ions

Con

stru

ctio

nof

Dum

my

Ato

ms

The

goal

ofde

finin

ghy

drog

enat

oms

asdu

mm

yat

oms

isto

rem

ove

allh

igh-

freq

uenc

yde

gree

sof

free

dom

from

them

.In

som

eca

ses

not

alld

egre

esof

free

dom

ofa

hydr

ogen

atom

shou

ldbe

rem

oved

,e.

g.in

the

case

ofhy

drox

ylor

amin

egr

oups

the

rota

tiona

lfr

eedo

mof

the

hydr

ogen

Page 130: CS USER AL Simulations - GROMACS

11

0C

ha

pte

r6

.S

pe

cialTo

pics

atom(s)

shouldbe

preserved.C

areshould

betaken

thatno

unwanted

correlationsare

introducedby

theconstruction

ofdum

my

atoms,

e.g.bond-angle

vibrationbetw

eenthe

constructingatom

scould

translateinto

hydrogenbond-length

vibration.A

dditionally,since

dumm

yatom

sare

bydefinition

mass-less,in

orderto

preservetotalsystem

mass,the

mass

ofeachhydrogen

atomthat

istreated

asdum

my

atomshould

beadded

tothe

bondedheavy

atom.

Takinginto

accounttheseconsiderations,the

hydrogenatom

sin

aprotein

naturallyfallinto

severalcategories,each

requiringa

differentapproach,seealso

Fig.

6.3:

•h

ydro

xyl(-O

H)

or

sulfh

ydryl

(-SH

)h

ydroge

n:T

heonly

internaldegree

offreedom

ina

hydroxylgroupthat

canbe

constrainedis

thebending

ofthe

C-O

-Hangle.

This

angleis

fixedby

definingan

additionalbondofappropriate

length,seeF

ig.6.3A

.This

removes

thehigh

frequencyangle

bending,butleavesthe

dihedralrotationalfreedom.

The

same

goesfor

asulfhydrylgroup.

Note

thatinthese

casesthe

hydrogenis

nottreatedas

adum

my

atom.

•sin

gle

am

ine

or

am

ide

(-NH

-)a

nd

aro

ma

tich

ydroge

ns

(-C

H-):

The

positionof

thesehy-

drogenscannot

beconstructed

froma

linearcom

binationof

bondvectors,

becauseof

theflexibility

oftheangle

between

theheavy

atoms.

Instead,the

hydrogenatom

ispositioned

ata

fixeddistance

fromthe

bondedheavy

atomon

aline

goingthrough

thebonded

heavyatom

anda

pointonthe

linethrough

bothsecond

bondedatom

s,seeF

ig.6.3B

.

•p

lan

ar

am

ine

(-NH

2 )h

ydroge

ns:T

hem

ethodused

forthe

singleam

idehydrogen

isnot

wellsuited

forplanar

amine

groups,because

nosuitable

two

heavyatom

scan

befound

todefine

thedirection

ofthehydrogen

atoms.

Instead,

thehydrogen

isconstructed

atafixed

distancefrom

thenitrogen

atom,w

itha

fixedangle

tothe

carbonatom

,inthe

planedefined

byone

oftheother

heavyatom

s,seeF

ig.6.3C

.

•a

min

eg

rou

p(u

mb

rella-N

H2

or

-NH

+3)

hyd

rogen

s:Am

inehydrogens

with

rotationalfree-dom

cannotbe

constructedas

dumm

yatom

sfrom

theheavy

atoms

theyare

connectedto,

sincethis

would

resultin

lossof

therotationalfreedom

ofthe

amine

group.To

preservethe

rotationalfreedomw

hilerem

ovingthe

hydrogenbond-angle

degreesof

freedom,

two

“dumm

ym

asses”are

constructedw

iththe

same

totalmass,m

omentofinertia

(forrotation

aroundtheC

-Nbond)

andcenter

ofm

assas

theam

inegroup.

These

dumm

ym

asseshave

nointeraction

with

anyother

atom,exceptfor

thefactthatthey

areconnected

tothe

carbonand

toeach

other,resultingin

arigid

triangle.F

romthese

threeparticles

thepositions

ofthenitrogen

andhydrogen

atoms

areconstructed

aslinearcom

binationsofthe

two

carbon-mass

vectorsand

theirouter

product,resultingin

anam

inegroup

with

rotationalfreedomintact,

butwithoutother

internaldegreesoffreedom

.S

eeF

ig.6.3D

.

6.2.2O

ut-of-planevibrations

inarom

aticgroups

The

planararrangem

entsin

theside

chainsof

thearom

aticresidues

lendsitself

perfectlyfor

adum

my-atom

construction,giving

aperfectly

planargroup

without

theinherently

instablecon-

straintsthatare

necessaryto

keepnorm

alatoms

ina

plane.T

hebasic

approachis

todefine

threeatom

sor

dumm

ym

assesw

ithconstraints

between

themto

fixthe

geometry

andcreate

therestof

theatom

sas

simple

dumm

ytype

3atom

s(see

sec.4.5)

fromthese

three.E

achof

thearom

aticresidues

requirea

differentapproach:

7.3

.R

un

Pa

ram

ete

rs1

19

7.3.5E

nergym

inimization

emtol:

(100.0)[kJ

mol −

1nm

−1]

them

inimization

isconverged

when

them

aximum

forceis

smaller

thanthis

value

emstep:

(0.01)[nm

]initialstep-size

nstcgsteep:(1000)

[steps]frequency

ofperforming

1steepestdescentstep

while

doingconjugate

gradientenergym

in-im

ization.

7.3.6S

hellMolecular

Dynam

ics

The

shellmoleculardynam

icsprogramxm

drunoptim

izesthe

positionsofthe

shellsatevery

time

stepuntileitherthe

RM

Sforce

onthe

shellsis

lessthan

emtol,ora

maxim

umnum

berofiterations(niter)

hasbeen

reached

emtol:

(100.0)[kJ

mol −

1nm

−1]

them

inimization

isconverged

when

them

aximum

forceis

smaller

thanthis

value.F

orshell

MD

thisvalue

shouldbe

1.0atm

ost,butsincethe

variableis

usedfor

energym

inimization

asw

ellthedefaultis

100.0.

niter:(20)m

aximum

number

ofiterationsfor

optimizing

theshellpositions.

7.3.7O

utputcontrol

nstxout:(100)

[steps]frequency

tow

ritecoordinates

tooutputtrajectory

file,thelastcoordinates

arealw

aysw

rit-ten

nstvout:(100)

[steps]frequency

tow

ritevelocities

tooutputtrajectory,the

lastvelocitiesare

always

written

nstfout:(0)

[steps]frequency

tow

riteforces

tooutputtrajectory.

nstlog:(100)

[steps]frequency

tow

riteenergies

tolog

file,thelastenergies

arealw

aysw

ritten

nstenergy:(100)

[steps]frequency

tow

riteenergies

toenergy

file,thelastenergies

arealw

aysw

ritten

nstxtcout:(0)

[steps]frequency

tow

ritecoordinates

toxtc

trajectory

xtcprecision:

(1000)[real]

precisionto

write

toxtc

trajectory

Page 131: CS USER AL Simulations - GROMACS

11

8C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

sdA

leap

-fro

gst

ocha

stic

dyna

mic

sin

tegr

ator

.T

hete

mpe

ratu

refo

rone

orm

ore

grou

psof

atom

s(tc

grps

)is

setw

ithre

ft

[K],

the

inve

rse

fric

tion

cons

tant

for

each

grou

pis

set

with

tau

t[p

s].

The

para

met

ertco

upli

sig

nore

d.T

hera

ndom

gene

rato

ris

initi

aliz

edw

ithld

seed

.

NO

TE

:tem

pera

ture

devi

atio

nsde

cay

twic

eas

fast

asw

itha

Ber

ends

enth

erm

osta

twith

the

sam

etau

t.

bdA

nE

uler

inte

grat

orfo

rB

row

nian

orpo

sitio

nLa

ngev

indy

nam

ics,

the

velo

city

isth

efo

rce

divi

ded

bya

fric

tion

coef

ficie

nt(bd

fric

[am

ups−

1])

plus

rand

omth

erm

alno

ise

(bd

tem

p[K

]).

Whe

nbd

fric

=0,

the

fric

tion

coef

ficie

ntfo

rea

chpa

rtic

leis

calc

u-la

ted

asm

ass/ta

ut,

asfo

rth

ein

tegr

ators

d.

The

rand

omge

nera

tor

isin

itial

ized

with

ldse

ed.

tinit:

(0)

[ps]

star

ting

time

for

your

run

(onl

ym

akes

sens

efo

rin

tegr

ator

sm

d,sd

and

bd

)

dt:

(0.0

01)

[ps]

time

step

for

inte

grat

ion

(onl

ym

akes

sens

efo

rin

tegr

ator

sm

d,sd

and

bd

)

nste

ps:

(1)

max

imum

num

ber

ofst

eps

toin

tegr

ate

nstc

omm

:(1

)[s

teps

]if

posi

tive:

freq

uenc

yfo

rce

nter

ofm

ass

mot

ion

rem

oval

ifne

gativ

e:fr

eque

ncy

for

cent

erof

mas

sm

otio

nan

dro

tatio

nalm

otio

nre

mov

al(s

houl

don

lybe

used

for

vacu

umsi

mul

atio

ns)

com

mgr

ps:

grou

p(s)

for

cent

erof

mas

sm

otio

nre

mov

al,

defa

ult

isth

ew

hole

syst

em,

rota

tion

rem

oval

can

only

bedo

neon

the

who

lesy

stem

7.3.

4La

ngev

indy

nam

ics

bdte

mp:

(300

)[K

]te

mpe

ratu

rein

Bro

wni

andy

nam

ics

run

(con

trol

sth

erm

alno

ise

leve

l).W

hen

bdfr

ic=

0,re

ft

isus

edin

stea

d.

bdfr

ic:

(0)

[am

ups

−1]

Bro

wni

andy

nam

ics

fric

tion

coef

ficie

nt.

Whe

nbdfr

ic=

0,th

efr

ictio

nco

effic

ient

for

each

part

icle

isca

lcul

ated

asm

ass/tau

t.

ldse

ed:

(199

3)[in

tege

r]us

edto

initi

aliz

era

ndom

gene

rato

rfo

rth

erm

alno

ise

for

stoc

hast

ican

dB

row

nian

dyna

m-

ics.

Whe

nld

seed

isse

tto

-1,

the

seed

isca

lcul

ated

as(tim

e()

+g

etp

id()

)%

10

00

00

0

6.3

.V

isco

sity

calc

ula

tion

11

1

ε

η

ζδ

ε

γ

ε

δε

δεδ γ

ζε

η

εδ

γ

Ph

eT

yrH

isT

rp

ζ

ε

ζ

εδ

γ

δδ

���������� ����������

���������� �������������������� ����������

���������� �������������� ����

���������� ����������

� ������������� ����������

���������� ����������

���������� ����������

���������� ����������

���������� ����������

���������� ����������

���������� ����������

Fig

ure

6.4:

The

diffe

rent

type

sof

dum

my

atom

cons

truc

tions

used

for

arom

atic

resi

dues

.T

heat

oms

used

inth

eco

nstr

uctio

nof

the

dum

my

atom

(s)

are

depi

cted

asbl

ack

circ

les,

dum

my

atom

sas

grey

ones

.H

ydro

gens

are

smal

ler

than

heav

yat

oms.

A:

phen

ylal

anin

e;B

:ty

rosi

ne(n

ote

that

the

hydr

oxyl

hydr

ogen

isno

tadu

mm

yat

om);C

:try

ptop

hane

;D:h

istid

ine.

•P

he

nyl

ala

nin

e:C

γ,C

ε1an

dCε2

are

kept

asno

rmal

atom

s,bu

twith

each

am

ass

ofon

eth

irdth

eto

talm

ass

ofth

eph

enyl

grou

p.S

eeF

ig.

6.3A

.

•Ty

rosi

ne

:The

ring

istr

eate

did

entic

alto

the

phen

ylal

anin

erin

g.A

dditi

onal

ly,

cons

trai

nts

are

defin

edbe

twee

nCε1

and

Cε2

and

Oη.

The

orig

inal

impr

oper

dihe

dral

angl

esw

illke

epbo

thtr

iang

les

(one

for

the

ring

and

one

withO

η)

ina

plan

e,bu

tdue

toth

ela

rger

mom

ents

ofin

ertia

this

cons

truc

tion

will

bem

uch

mor

est

able

.T

hebo

ndan

gle

inth

ehy

drox

ylgr

oup

will

beco

nstr

aine

dby

aco

nstr

aint

betw

een

and

Hη,n

ote

that

the

hydr

ogen

isno

ttre

ated

asa

dum

my

atom

.S

eeF

ig.

6.3B

.

•T

ryp

top

ha

ne

:Cβ

iske

ptas

ano

rmal

atom

and

two

dum

my

mas

ses

are

crea

ted

atth

ece

nter

ofm

ass

ofea

chof

the

rings

,eac

hw

itha

mas

seq

ualt

oth

eto

talm

ass

ofth

ere

spec

tive

ring

(Cδ2

and

Cε2

are

each

coun

ted

half

for

each

ring)

.T

his

keep

sth

eov

eral

lcen

ter

ofm

ass

and

the

mom

ento

fine

rtia

alm

ost(

butn

otqu

ite)

equa

lto

wha

titw

as.

See

Fig

.6.

3C.

•H

istid

ine

:C

γ,

Cε1

and

Nε2

are

kept

asno

rmal

atom

s,bu

tw

ithm

asse

sre

dist

ribut

edsu

chth

atth

ece

nter

ofm

ass

ofth

erin

gis

pres

erve

d.S

eeF

ig.

6.3D

.

6.3

Vis

cosi

tyca

lcul

atio

n

The

shea

rvi

scos

ityis

apr

oper

tyof

liqui

dw

hich

can

bede

term

ined

easi

lyby

expe

rimen

t.It

isus

eful

for

para

met

eriz

ing

the

forc

efiel

d,be

caus

eit

isa

kine

ticpr

oper

ty,w

hile

mos

toth

erpr

oper

-tie

sw

hich

are

used

for

para

met

eriz

atio

nar

eth

erm

odyn

amic

.T

hevi

scos

ityis

also

anim

port

ant

prop

erty

,si

nce

itis

ofin

fluen

ceon

the

rate

sof

conf

orm

atio

nalc

hang

esof

mol

ecul

esso

lvat

edin

the

liqui

d.

The

visc

osity

can

beca

lcul

ated

from

aneq

uilib

rium

sim

ulat

ion

usin

gan

Ein

stei

nre

latio

n:

η=

1 2V

kBT

lim t→∞

d dt

⟨ ( ∫t 0

+t

t 0P

xz(t′ )

dt′) 2⟩

t 0

(6.1

)

Thi

sca

nbe

done

withg

en

erg

y.

Thi

sm

etho

dco

nver

ges

very

slow

ly.

Ana

nose

cond

sim

ulat

ion

mig

htno

tbe

long

enou

ghfo

ran

accu

rate

dete

rmin

inat

ion

ofth

evi

scoi

ty.

The

resu

ltis

very

depe

nden

ton

the

trea

tmen

tof

the

elec

tros

tatic

s.U

sing

a(s

hort

)cu

t-of

fre

sults

inla

rge

nois

eon

Page 132: CS USER AL Simulations - GROMACS

11

2C

ha

pte

r6

.S

pe

cialTo

pics

theoff-diagonal

pressureelem

ents,w

hichcan

increasethe

calculatedviscosity

byan

orderof

magnitude.

GR

OM

AC

Salso

hasa

non-equilibriumm

ethodfor

determining

theviscosity.

This

makes

useof

thefact

thatenergy,

which

isfed

intosystem

byexternalforces,

isdissipated

throughviscous

friction.T

hegenerated

heatisrem

ovedby

couplingto

aheatbath.

For

aN

ewtonian

liquidadding

asm

allforcew

illresultina

velocitygradientaccording

tothe

following

equation:

ax (z)

+ηρ

∂2v

x (z)∂z2

=0

(6.2)

herew

ehave

appliedan

accelerationa

x (z)in

thex

-direction,w

hichis

afunction

ofthez-

coordinate.In

GR

OM

AC

Sthe

accelerationprofile

is:

ax (z)

=A

cos (2πz

lz )(6.3)

where

lzis

theheightofthe

box.T

hegenerated

velocityprofile

is:

vx (z)

=V

cos (2πz

lz )(6.4)

V=Aρη (

lz2π )

2

(6.5)

The

viscositycan

becalculated

fromAand

V:

η=AVρ (

lz2π )

2

(6.6)

Inthe

simulationV

isdefined

as:V=

N∑i=1

mi v

i,x 2cos (

2πz

lz )N∑i=

1

mi

(6.7)

The

generatedvelocity

profileis

notcoupledto

theheatbath,also

thevelocity

profileis

excludedfrom

thekinetic

energy.O

new

ouldlikeV

tobe

aslarge

aspossible

togetgood

statistics.H

owever

theshear

rateshould

notbeso

highthatthe

systemgets

toofar

fromequilibrium

.T

hem

aximum

shearrate

occursw

herethe

cosineis

zero,therate

is:

shm

ax

=m

axz ∣∣∣∣ ∂vx (z)∂z ∣∣∣∣ =

Aρη

lz2π(6.8)

For

asim

ulationw

ith:η=

10−

3[kg

m−

1s −

1],ρ

=10

3[kg

m−

3]and

lz=

2π[nm

],shm

ax

=1

[psnm

−1]A

.T

hisshear

rateshould

besm

allerthan

oneover

thelongestcorrelation

time

inthe

system.

Form

ostliquidsthis

willbe

therotation

correlationtim

e,which

isaround

10picoseconds.

Inthis

caseAshould

besm

allerthan

0.1[nm

ps−

2].W

henthe

shearrate

istoo

high,theobserved

viscosityw

illbetoo

low.

BecauseV

isproportionalto

thesquare

ofthe

boxheight,

theoptim

al

7.3

.R

un

Pa

ram

ete

rs1

17

7.3R

unP

arameters

7.3.1G

eneral

Default

valuesare

givenin

parentheses.T

hefirst

optionis

always

thedefault

option.U

nitsare

givenin

squarebrackets

The

differencebetw

eena

dashand

anunderscore

isignored.

Asam

ple.m

dp

fileis

available.T

hisshould

beappropriate

tostarta

normalsim

ulation.E

ditittosuityour

specificneeds

anddesires.

7.3.2P

reprocessing

title:this

isredundant,so

youcan

typeanything

youw

ant

cpp:(/lib/cpp)your

preprocessor

include:directories

toinclude

inyour

topology.form

at:-I/h

om

e/jo

hn

/my

lib-I../m

ore

lib

define:()

definesto

passto

thepreprocessor,defaultis

nodefines.

Youcan

useany

definesto

controloptions

inyour

customized

topologyfiles.

Options

thatarealready

availableby

defaultare:

-DF

LEX

SP

CW

illtellgrompp

toinclude

FLE

XS

PC

instead

ofSP

Cinto

yourtopology,this

isnec-

essaryto

makeconjugate

gradientwork

andw

illallowsteepestdescentto

minim

izefurther.

-DP

OS

RE

Willtellgrom

ppto

includeposre.itp

intoyour

topology,usedfor

positionrestraints.

7.3.3R

uncontrol

integrator:

md

Aleap-frog

algorithmfor

integratingN

ewton’s

equations.

steepAsteepest

descentalgorithm

forenergy

minim

ization.T

hem

aximum

stepsize

isem

step[nm],the

toleranceisem

tol[kJm

ol −1

nm−

1].

cgA

conjugategradient

algorithmfor

energym

inimization,

thetolerance

isem

tol[kJm

ol −1

nm−

1].C

Gis

more

efficientwhen

asteepestdescentstep

isdone

everyonce

ina

while,this

isdeterm

inedbynstcgsteep.

Page 133: CS USER AL Simulations - GROMACS

11

6C

ha

pte

r7

.R

un

pa

ram

ete

rsa

nd

Pro

gra

ms

Def

ault

Def

ault

Nam

eE

xt.

Type

Opt

ion

Des

crip

tion

ato

mtp

.atp

Asc

Ato

mty

pefil

eus

edby

pdb2

gmx

eiw

it.b

rkA

sc-f

Bro

okha

ven

data

bank

file

nn

nic

e.d

at

Asc

Gen

eric

data

file

use

r.d

lgA

scD

ialo

gB

oxda

tafo

rng

mx

sam

.ed

iA

scE

Dsa

mpl

ing

inpu

tsa

m.e

do

Asc

ED

sam

plin

gou

tput

en

er.

ed

rG

ener

icen

ergy

:edr

en

ee

ne

r.e

dr

xdr

Ene

rgy

file

inpo

rtab

lexd

rfo

rmat

en

er.

en

eB

inE

nerg

yfil

ee

iwit.

en

tA

sc-f

Ent

ryin

the

prot

ein

date

bank

plo

t.e

ps

Asc

Enc

apsu

late

dP

ostS

crip

t(tm

)fil

eg

tra

j.g8

7A

scG

rom

os-8

7A

SC

IItr

ajec

tory

form

atco

nf.g

96

Asc

-cC

oord

inat

efil

ein

Gro

mos

-96

form

atco

nf.g

ro-c

Gen

eric

stru

ctur

e:gro

g9

6p

db

tpr

tpb

tpa

ou

t.g

ro-o

Gen

eric

stru

ctur

e:gro

g9

6p

db

con

f.g

roA

sc-c

Coo

rdin

ate

file

inG

rom

os-8

7fo

rmat

po

lar.

hd

bA

scH

ydro

gen

data

base

top

inc.

itpA

scIn

clud

efil

efo

rto

polo

gyru

n.lo

gA

sc-l

Log

file

ps.

m2

pA

scIn

putfi

lefo

rm

at2p

sss

.ma

pA

scF

ileth

atm

aps

mat

rixda

tato

colo

rsss

.ma

tA

scM

atrix

Dat

afil

eg

rom

pp

.md

pA

sc-f

grom

ppin

putfi

lew

ithM

Dpa

ram

eter

sh

ess

ian

.mtx

Bin

-mH

essi

anm

atrix

ind

ex.

nd

xA

sc-n

Inde

xfil

eh

ello

.ou

tA

sc-o

Gen

eric

outp

utfil

ee

iwit.

pd

bA

sc-f

Pro

tein

data

bank

file

pu

ll.p

do

Asc

Pul

ldat

aou

tput

pu

ll.p

pa

Asc

Pul

lpar

amet

ers

resi

du

e.r

tpA

scR

esid

ueTy

pefil

eus

edby

pdb2

gmx

do

c.te

xA

sc-o

LaTe

Xfil

eto

po

l.to

pA

sc-p

Topo

logy

file

top

ol.t

pa

Asc

-sA

scii

run

inpu

tfile

top

ol.t

pb

Bin

-sB

inar

yru

nin

putfi

leto

po

l.tp

r-s

Gen

eric

run

inpu

t:tp

rtp

btp

ato

po

l.tp

r-s

Str

uctu

re+

mas

s(db

):tpr

tpb

tpa

gro

g9

6p

db

top

ol.t

pr

xdr

-sP

orta

ble

xdr

run

inpu

tfile

tra

j.trj

Bin

Tra

ject

ory

file

(cpu

spec

ific)

tra

j.trr

Ful

lpre

cisi

ontr

ajec

tory

:trr

trj

tra

j.trr

xdr

Tra

ject

ory

inpo

rtab

lexd

rfo

rmat

roo

t.xp

mA

scX

Pix

Map

com

patib

lem

atrix

file

tra

j.xtc

-fG

ener

ictr

ajec

tory

:xtc

trr

trj

gro

g9

6p

db

tra

j.xtc

xdr

Com

pres

sed

traj

ecto

ry(p

orta

ble

xdr

form

at)

gra

ph

.xvg

Asc

-oxv

gr/x

mgr

file

Tabl

e7.

1:T

heG

RO

MA

CS

file

type

san

dho

wth

eyca

nbe

used

for

inpu

t/out

put.

6.4

.U

ser

spe

cifie

dp

ote

ntia

lfu

nct

ion

s1

13

box

isel

onga

ted

inth

ez-di

rect

ion.

Inge

nera

lasi

mul

atio

nle

ngth

of10

0pi

cose

cond

sis

enou

ghto

obta

inan

accu

rate

valu

efo

rth

evi

scos

ity.

The

heat

gene

rate

dby

the

visc

ous

fric

tion

isre

mov

edby

coup

ling

toa

heat

bath

.B

ecau

seth

isco

uplin

gis

not

inst

anta

neou

sth

ere

alte

mpe

ratu

reof

the

liqui

dw

illbe

slig

htly

low

erth

anth

eob

serv

edte

mpe

ratu

re.

Ber

ends

ende

rived

this

tem

pera

ture

shift

[20

],w

hich

can

bew

ritte

nin

term

sof

the

shea

rra

teas

:T

s=

ητ

2ρC

vsh

2 max

(6.9

)

whe

reτ

isth

eco

uplin

gtim

efo

rth

eB

eren

dsen

ther

mos

tat

and

Cv

isth

ehe

atca

paci

ty.

Usi

ngth

eva

lues

ofth

eex

ampl

eab

ove,τ

=10−

13

[s]

andC

v=

2·1

03[J

kg−

1K−

1],

we

get:T

s=

25[K

ps−

2]s

h2 max.

Whe

nw

ew

antt

hesh

ear

rate

tobe

smal

ler

than

1/10

[ps−

1],T

sis

smal

ler

than

0.25

[K],

whi

chis

negl

igib

le.

Not

eth

atth

esy

stem

has

tobu

ildup

the

velo

city

profi

lew

hen

star

ting

from

aneq

uilib

rium

stat

e.T

his

build

-up

time

isof

the

orde

rof

the

corr

elat

ion

time

ofth

eliq

uid.

Two

quan

titie

sar

ew

ritte

nto

the

ener

gyfil

e,al

ong

with

thei

rave

rage

san

dflu

ctua

tions

:V

and1/η

asob

tain

edfr

om(6.6)

.

6.4

Use

rsp

ecifi

edpo

tent

ialf

unct

ions

You

can

also

use

your

own

pote

ntia

lfun

ctio

nsw

ithou

tedi

ting

the

GR

OM

AC

Sco

de.

The

pote

ntia

lfu

nctio

nsh

ould

beac

cord

ing

toth

efo

llow

ing

equa

tion

V(r

ij)

=q iq j

4πε 0f(r

ij)+C

6g(r

ij)+C

12h(r

ij)

(6.1

0)

with

f,g,h

user

defin

edfu

nctio

ns.

Not

eth

atif

g(r)

repr

esen

tsa

norm

aldi

sper

sion

inte

ract

ion,

g(r)

shou

ldbe<

0.C

6,

C12

and

the

char

ges

are

read

from

the

topo

logy

.A

lso

note

that

com

bina

tion

rule

sar

eon

lysu

ppor

ted

for

Lenn

ard

Jone

san

dB

ucki

ngha

m,

and

that

your

tabl

essh

ould

mat

chth

epa

ram

eter

sin

the

bina

ryto

polo

gy.

Whe

nyo

uad

dth

efo

llow

ing

lines

inyo

ur.md

pfil

e:

rlis

t=

1.0

cou

lom

bty

pe

=U

ser

rco

ulo

mb

=1

.0vd

wty

pe

=U

ser

rvd

w=

1.0

the

MD

prog

ram

will

read

asi

ngle

file

(nam

eca

nbe

chan

ged

with

optio

n-t

ab

le)

with

seve

nco

lum

nsof

tabl

elo

okup

data

inth

eor

der:

x,f(

x),f

”(x)

,g(x

),g”

(x),

h(x)

,h”(

x).

The

xsh

ould

run

from

0to

r c+0.

5,w

itha

spac

ing

of0.

002

nmw

hen

you

run

insi

ngle

prec

isio

n,or

0.00

05w

hen

you

run

indo

uble

prec

isio

n.In

this

cont

extr

cde

note

sth

em

axim

umof

the

two

cut-

offsrvd

wan

drc

ou

lom

b(s

eeab

ove)

.T

hese

varia

bles

need

notb

eth

esa

me

(and

need

notb

e1.

0ei

ther

).S

ome

func

tions

used

for

pote

ntia

lsco

ntai

na

sing

ular

ityat

x=

0,bu

tsin

ceat

oms

are

norm

ally

notc

lose

rto

each

othe

rth

an0.

1nm

,the

func

tion

valu

eat

x=

0is

noti

mpo

rtan

tand

the

tabl

esca

nbe

star

ted

ate.

g.0.

02nm

.F

inal

ly,

itis

also

poss

ible

toco

mbi

nea

stan

dard

Cou

lom

bw

itha

mod

ified

LJ

Page 134: CS USER AL Simulations - GROMACS

11

4C

ha

pte

r6

.S

pe

cialTo

pics

potential(orvice

versa).O

nethen

specifiese.g.

coulombtype

=C

ut-offorcoulom

btype=

PM

E,

combined

with

vdwtype

=U

ser.T

hetable

filem

ustalw

ayscontain

the7

columns

however,

andm

eaningfuldata(i.e.

notzeroes)m

ustbeentered

inallcolum

ns.A

number

ofpre-builttablefiles

canbe

foundin

theG

MX

LIBdirectory,

for6-8,

6-9,6-10,

6-11,6-12

LennardJones

potentialscom

binedw

itha

normalC

oulomb.

6.5R

unningG

RO

MA

CS

inparallel

Ifyou

haveinstalled

theM

PI

(Message

Passing

Interface)on

yourcom

puter(s)you

cancom

pileG

RO

MA

CS

with

thislibrary

torun

simulations

inparallel.

Allsupercom

putersare

shippedw

ithM

PIlibraries

optimized

forthatparticular

platform,and

ifyouare

usinga

clusterofw

orkstationsthere

areseveralgood

freeM

PIim

plementations.

Youcan

findupdated

linksto

theseon

thegro-

macs

homepagew

ww

.gromacs.org.O

nceyou

havean

MP

Ilibraryinstalled

it’strivialto

compile

GR

OM

AC

Sw

ithM

PI

support:Just

setthe

option--en

ab

le-m

pi

tothe

configurescript

andrecom

pile.(B

utdon’t

forgetto

make

distcleanbefore

runningconfigure

ifyou

havepreviously

compiled

with

adifferentconfiguration.)

Ifyouare

usinga

supercomputer

youm

ightalsow

anttoturn

ofthedefaultnicing

ofthem

drunprocess

with

the--d

isab

le-n

iceoption.

There

isusually

aprogram

calledmpiru

nw

ithw

hichyou

canfire

upthe

parallelprocesses.A

typicalcomm

andline

lookslike:

%m

piru

n-p

go

ofu

s,do

ofu

s,fred

10

md

run

-sto

po

l-v

-N3

0this

runson

eachofthe

machines

goofus,doofus,fredw

ith10

processeson

each1.

Ifyou

havea

singlem

achinew

ithm

ultipleprocessors

youdon’t

haveto

usethe

mp

irun

com-

mand,butyou

cando

with

anextra

optionto

md

run

:%

md

run

-np

8-s

top

ol

-v-N

8In

thisexam

pleM

PI

readsthe

firstoption

fromthe

comm

andline.

Sincem

dru

nalso

wants

toknow

thenum

berofprocesses

youhave

totype

ittwice.

Check

yourlocalm

anuals(or

onlinem

anual)for

exactdetailsofyour

MP

Iimplem

entation.

Ifyouare

interestedin

programm

ingM

PIyourself,you

canfind

manuals

andreference

litteratureon

thew

ebatw

ww

.mcs.anl.gov/m

pi/index.html

.

1Exam

pletaken

fromS

iliconG

raphicsm

anual

Chapter

7

Run

parameters

andP

rograms

7.1O

nlineand

htmlm

anuals

Allthe

information

inthis

chaptercan

alsobe

foundin

HT

ML

format

inyour

GR

OM

AC

Sdata

directory.T

hepath

dependson

where

yourfiles

areinstalled,butthe

defaultlocationis

/usr/lo

cal/g

rom

acs/sh

are

/htm

l/on

line

.htm

lO

r,ifyouinstalled

fromLinux

packagesitcan

befound

as/u

sr/loca

l/sha

re/g

rom

acs/h

tml/o

nlin

e.h

tml

Youcan

alsouse

theonline

fromour

web

site,w

ww

.gromacs.org/docum

entation/reference3.0/online.htm

l

Inaddition,

we

installstandard

UN

IXm

anualsfor

allthe

programs.

Ifyou

havesourced

theG

MX

RCscriptin

theG

RO

MA

CS

binarydirectory

foryour

hosttheyshould

alreadybe

presentinyour$

MA

NP

AT

H,andyou

shouldbe

ableto

typee.g.m

an

gro

mp

p.

The

programm

anualpagescan

alsobe

foundin

Appendix

Ein

thism

anual.

7.2F

iletypes

Table7.1

liststhe

filetypes

usedby

GR

OM

AC

Salong

with

ashortdescription,and

youcan

finda

more

detaildescriptionfor

eachfile

inyour

HT

ML

reference,orin

ouronline

version.

GR

OM

AC

Sfiles

written

inxdr

format

canbe

readon

anyarchitecture

with

GR

OM

AC

Sversion

1.6orlaterifthe

configurationscriptfound

theX

DR

librarieson

yoursystem.

They

shouldalw

aysbe

presentonU

NIX

sincethey

arenecessary

forthe

network

filesystem

support.