Evaluating the Solution from MrBUMP and Balbes

Ronan Keegan, Fei Long, Martyn Winn, Garib Murshudov

STFC Daresbury Laboratory&

York Structural Biology Laboratory, University of York

PDB Depositions (1999-2009)

1999 2009

What are MrBUMP and Balbes?

• Automated molecular replacement (MR) and beyond....

• MrBUMP: – Brute force approach - try a variety of search

models prepared in several different ways in MR

• Balbes: – Identifies the best possible search model to be

used in MR by using a specially grafted version of the PDB optimised for use in MR

MrBUMP

•An automation framework for Molecular Replacement.•Particular emphasis on generating a variety of search models.

•In favourable cases, gives “one-button” solution•In Complicated Cases, will suggest likely search models for manual investigation (lead generation)

• Wraps Phaser and/or Molrep.• Uses a variety of helper applications (e.g. Chainsaw) and bioinformatics tools (e.g. Fasta, Mafft) to generate search models• Makes use of up-to-date on-line databases (e.g. PDB, Scop)

`

`

`

`Target MTZ &

Sequence

TargetDetails

TemplateSearch

ModelPreparation

Molecular Replacement& Refinement

Check scores and exit or select the next model

The Pipeline

Phase Improvement

N templates

N x M models

Search for model templates• FASTA search of PDB– Sequence based search using sequence of target structure

All of the resulting PDB id codes are added to a list

These structures are calledmodel templates

• Other templates from:o SSM search using top hit from the FASTA search o Can add additional PDB id codes to the listo Can add local PDB files

Multiple Alignment steptarget

modeltemplates

pairwisealignment(used in Chainsaw)Jalview 2.08.1 Barton group, Dundee

currently support ClustalW, MAFFT, probcons or T-coffee for multiple alignment

Model template scoring: score = sequence identity X alignment quality

template chains

XSCOP

Domainse.g. if relative domain motion

PQS/PISA

superpose

EnsemblesCreate ensemble of top search models, for use in additional run of Phaser.Need to be similar in MW and rmsd

MultimersBetter signal-to-noise ratio than monomer, if assembly is correct for the target.

Domain 1

Domain 2

Search Model Preparation

Search models prepared in four ways:PDBclip

– original PDB with waters removed, most probable conformations selected and format tidied (e.g. chain ID added)

Molrep – Molrep model preparation function which

aligns the template sequence with the target sequence and prunes the non-conserved side chains accordingly.

Chainsaw – Can be given any alignment between the

target and template sequences. Non-conserved residues are pruned back to the gamma atom.

Polyalanine– Created by excluding all of the side chain

atoms beyond the CB atom using the Pdbset program

more side chain truncation

deal with deletions

Running MR• For each search model, MR done with Molrep or Phaser or both.• MR programs run mostly with defaults• MrBUMP provides LABIN columns, MW of target, sequence identity of

search model, number of copies to search for, number of clashes tolerated

Molecular replacement step

Molecular replacement step

• MR output– MR scores and un-refined

models available for later inspection

assess quality of solution, extent of model bias

– MrBUMP doesn’t use MR scores, but checks for output file with positioned model, and passes to refinement step

Testing enantiomorphic spacegroups

– 11 pairs of enantiomorphic spacegroups containing screw axes of opposite handedness, e.g. P41 and P43)

– usually both need to be tested in MR– correct spacegroup indicated by TF and packing

– MrBUMP can test both in Molrep and/or Phaser.– For each search model, best MR results used to fix

spacegroup for subsequent steps.– Discrimination good for good search model + correct MR

solution

Inclusion of fixed models• MrBUMP will now accept one or more positioned models.• These are included as fixed models in all MR jobs.

• Thus, solve complexes through consecutive runs of MrBUMP.• Automation of this in progress ....

otherwise

final Rfree < 0.48 orfinal Rfree < 0.52 and dropped by 5%

• The resulting models from molecular replacement are passed to Refmac for restrained refinement.

• The change in the Rfree value during refinement is used as rough estimate of how good the resulting model is.

final Rfree < 0.35 or final Rfree < 0.5 and dropped by 20%

“success”

“marginal”

“poor”

conservative .....

Restrained refinement step

Refinement results

Summarised results..

Best search model so far and file location for this model

List of sorted results so far

Example (thanks to Elien Vandermarliere)

Target is an arabinofuranosidaseData to 1.55Å in P212121

Small C domain (144 res) solved with 34% seq ident model(1w9t_B_MOLREP best out of 4 solutions)

With C domain solution fixed, large N domain (345 res) solved with 28% seq ident model(1gyh_C_CHNSAW best out of 7 solutions)

Acorn: CC increases from 0.04 to 0.18This step part of MrBUMPARP/wARP then builds 457/493 residues to R/Rfree 0.185/0.225

MrBUMP outputLog file gives summary of models tried and results of MR

• May get several putative solutions• Ease of subsequent model re-building, model completion may depend on choice of solution• Worth checking “poor” solutions

Top solution available from ccp4i

Using MrBUMP

• Part of CCP4 suite• CCP4i GUI

(Balbes)

http://www.ysbl.york.ac.uk/~fei/balbes/index.html

Fei Long, Alexei Vagin, Paul Young, Garib MurshudovYSBL, York University

Balbes• Balbes uses a reorganised version of the PDB

database customised for use in MR.• Has its own classification of domains most suitable

for use in molecular replacement• Database also includes classification of possible

oligomers based on the basic structures. • Built upon Molrep for molecular replacement and

Refmac for doing the refinement

Balbes Database• Cut down version of the PDB keeping only single copies of unique

folds.• Where 2 entries have sequence ID > 80 % and rmsd < 1 angstrom,

the entry with the highest resolution is retained• More than 20000 entries• More than 23000 domains classified• For each sequence additional information is stored including

secondary structure information, number of domains and potential to form multimers

• Flexible loops are cut from the stored structures• Multimers are classified with information from the EBI PISA

service

Balbes Workflow

Assemblies

• Fully automatic support for handling assemblies

• Simply provide additional sequences in input sequence file for each additional component in the target data

• Balbes will first look for assemblies in its database containing all of the sequences but if not find it will look for subsets of those sequences

Using Balbes

• Through CCP4i:– included in Linux and Mac OSX releases– Simple CCP4i interface - provide sequence and structure

factor file (MTZ)– Enter multiple sequences in input file for complexes

Using Balbes

• Through the YSBL Web Server:– Balbes is one of

several programs available to use via the web at the York University YSBL Web Server

– Create an account for access

http://www.ysbl.york.ac.uk/YSBLPrograms/index.jsp

Balbes Web Server

• Enter MTZ and sequence information as a file or paste sequence

• Option to check all related space groups

• Option to submit resulting map to ARP/wARP server for final model building

Web Server Results

• Summary of processing for each spacegroup

• Final best result highlighted• For each spacegroup log file

and all output files are made available for download (5 days)

• If user opted to use ARP/wARP server a link to the ARP/wARP results is provided

Balbes Output

• Spacegroup specific output:– Download files– Main summary file

showing results of MR and refinement for each template model that was used.

– Q value scoring

Balbes – the statistics

• In 2006, 67% of structures deposited in the PDB where solved by MR

• Balbes improved this to 80% using– A better organised database (for MR)– A better choice of protocols for tackling the MR problem– Improved algorithms in MR and refinement programs

Balbes Usage

• 90 structures in PDB referencing use of Balbes as part of structure solution

• 35 citations • More than 5000 users of web server version

of program• ~ 150 users per month

Summary• MrBUMP:

– Brute force approach, try everything– May give lead in borderline cases as well automating straight-forward

cases– All results are easily accessible and summarised

• Balbes:– Efficient and quick at solving structures that have reasonable

homologues– Deals with assemblies/complexes fully automatically– Combined with ARP/wARP it can take structure nearly all the way to

completion • In conclusion:

– Try Both!! Compute cycles are very cheap.

Acknowledgements

• Fei Long, Alexei Vagin, Paul Young, Andrey Lebedev, Garib Murshudov – Balbes and Refmac

• Martyn Winn, MrBUMP• Airlie McCoy, Randy Read – Phaser• Alexi Vagin – Molrep• Norman Stein – Chainsaw and Ctruncate• CCP4 Core team – Support and Testing

Downloading Balbes Database

– Balbes database available as a separate download (1.6GB)

PDB Statistics• The number of structures in the PDB is

increasing rapidly year on year• By the end of 2009 there were > 62000

structures deposited• 7449 structures were deposited in 2009 (12 %

of the total)• MrBUMP and Balbes seek to exploit this wealth

of data to improve the success rate of molecular replacement