Drug Discovery Grid-- A real grid application

Zhang Wenju, Shen Jianhua

Shanghai Institute of Materia Medica, CASShanghai Jiaotong University Jiangnan Institute of ComputingThe University of Hong Kong

Agenda

1. DDGrid Introduction

2. DDGrid Architecture

3. DDGrid Application

4. DDGrid Demo

Background

Large-scale High-throughput Virtual Screening

in SilicoThe computational analysis of chemical databases to identify compounds appropriate for a given biological receptor

in VitroIdentification of new compounds showing some activity against a target biological receptor, and the progressive optimization of these leads to yield a compound with improved potency and physicochemical properties in vitro

in Vivoeventually, improved efficacy, pharmacokinetics, and toxicological profiles in vivo.

Process of Drug Discovery and Design

2-3 2-3 yearsyears

3-4 3-4 yearsyears

Random Screening Random Screening 10, 000 ~ 20, 000 10, 000 ~ 20, 000

CompoundsCompounds

Random Screening Random Screening 10, 000 ~ 20, 000 10, 000 ~ 20, 000

CompoundsCompounds Drug CandidateDrug CandidateDrug CandidateDrug Candidate Pre-clinicPre-clinicPre-clinicPre-clinic

ClinicClinic(phase I, II, III)(phase I, II, III)

ClinicClinic(phase I, II, III)(phase I, II, III)

MarketMarketMarketMarket

2-3 2-3 yearsyears 2-3 2-3 yearsyears

Time: 10-12 yearsTime: 10-12 yearsMoney: several billion dollarsMoney: several billion dollars

ComputerComputer-Aid Drug -Aid Drug

DesignDesign

Leads and Opt.Leads and Opt.Leads and Opt.Leads and Opt.

DDGrid overview

◆ Drug Discovery Grid project aims to build a collaboration platform for drug discovery using the state-of-the-art grid computing technology.

◆ This project intends to solve large-scale computation and data intensive scientific applications in the fields of medicine chemistry and molecular biology with the help of grid middleware developed by our team.

◆ Over one million compounds database with 3-D structure and physicochemical properties are also provided to identify potential drug candidates. Users also can build and maintain their own customized ligand database to share in this grid platform.

DDGrid Architecture

Internet

Global Server

Slave Server

User User User

Internet

Slave Server Slave Server

DDGrid Architecture

Internet

Global Server

Slave Server

User 终端 终端

Internet

Slave Server Slave Server

Resource monitoring, job submit and monitor, input and parameter, result view and download through Web Portal

Resource monitoring, job submit and monitor, input and parameter, result view and download through Web Portal

DDGrid Architecture

Internet

Global Server

Slave Server

User User User

Internet

子服务器

•U

ser interface

•R

esources manag.

•Job subm

it and mon.

•K

ey and cert manag.

•R

esult analysis

•G

lobal scheduling

•visisualiszation

•D

istributed CD

B

DDGrid Architecture

Internet

主服务器

Slave server

User User User

Internet

slave slave

•Local job manag.•Local res. manag.•Local CDB manag.•Data en-decrypt•Local result

assimilate

•Local job manag.•Local res. manag.•Local CDB manag.•Data en-decrypt•Local result

assimilate

DDGrid Workflow

Job Submit

Global Server (Monitoring, Work Pool, Resource Manag., Assimilate of Result)

ID and Result Return

Slave Server (Local Resource Manag., Monitoring, Local Work Pool, Assimilate of Result)

Return of Result, New job requestJob Dispatch

Computational Client (Docking)

Job DispatchReturn of Result, New job request

xmlxml

DDGrid security

1. PKI-based security2. All the sites involved should hold a certification issued by our CA3. All the databases deployed and results are encrypted4. All the message passing are SSL/TLS-enabled

DDGrid Web Portal

Test Case 1

Virtual Screening from 20,000 compounds

Involved Sites:

Shanghai Inst. of M. M. (SIMM) Alpha Cluster (32CPU)

Beijing Mol. Ltd. Sunway Cluster (224CPU)

The Univ. of Hong KongGideon Cluster (16CPU)

Shanghai SuperComp. Centre Dawning 4000A

Dalian Univ. of Tech. Dawning 4000A

London e-Science Centre Mars Cluster

Time consumed:

5946 sec （ appr. 99 min）

Data Sets (CDB)：Specs

Job scheduling

Visualisation of Docking Result

DDGrid message passing

<scheduler_request> <authenticator>3333</authenticator> <hostid>102</hostid> <rpc_seqno>2401</rpc_seqno> <platform_name>i686-pc-linux-gnu</platform_name> <core_client_major_version>2</core_client_major_version> <core_client_minor_version>19</core_client_minor_version> <idle_ncpu>16</idle_ncpu> <project_disk_usage>5315768.000000</project_disk_usage> <total_disk_usage>68417940.000000</total_disk_usage> <code_sign_key> … </code_sign_key> <projects> <project> <master_url>http://www.ddgrid.ac.cn/ddg/</master_url> <resource_share>100.000000</resource_share> </project> </projects> <result> … </result> … <host_info> … </host_info></scheduler_request>

DDGrid message passing

<scheduler_reply> <message priority="low">No work available</message> <project_name>Ddg</project_name> <user_name>sss</user_name> <code_sign_key> … </code_sign_key> … <workunit> … </workunit> <preferences> <low_water_days>1.2</low_water_days> <high_water_days>2.5</high_water_days> <disk_max_used_gb>0.4</disk_max_used_gb> <disk_max_used_pct>50</disk_max_used_pct> <disk_min_free_gb>0.4</disk_min_free_gb> … </preferences> …</scheduler_reply>

DDGrid message passing

<workunit> <file_info> <number>0</number> </file_info> <file_info> <number>1</number> </file_info> <file_info> <number>2</number> </file_info> … <file_ref> <file_number>0</file_number> <open_name>tabfile</open_name> </file_ref> <file_ref> <file_number>1</file_number> <open_name>infile</open_name> </file_ref> <file_ref> <file_number>2</file_number> <open_name>sphfile</open_name> </file_ref> <command_line>-business</command_line></workunit>

DDGrid message passing

…<project> <scheduler_url>http://www.ddgrid.ac.cn/ddg_cgi/cgi</scheduler_url> <master_url>http://www.ddgrid.ac.cn/ddg/</master_url> <project_name>Ddg</project_name></project><app> <name>gridapp</name></app><file_info> <name>gridapp/gridapp_2.19_i686-pc-linux-gnu</name> <nbytes>260754.000000</nbytes> <max_nbytes>0.000000</max_nbytes> <executable/> <signature_required/> <file_signature> … </file_signature> <url>http://www.ddgrid.ac.cn/ddg/download/gridapp_2.19_i686-pc-linux-gnu</url></file_info><file_info> … </file_info>…

DDGrid Resources

Computational and Data Resources Integration

Resources aggregatedSIMM Sunway 32A Cluster

Beijing Molecule Inc. Sunway 256P ClusterHKU Gideon 300 ClusterSSC Dawning 4000ALeSC Mars Cluster (Test only)Singapore Poly-tech Univ.Dalian Univ. of TechnologyShanghai Jiaotong Univ.

Heterogeneous resourcesOS: IRIX, Digital Unix, Linux(IA32, x86_64)CPU ： R12000, Alpha, Pentium, AMD

DDGrid Resources

DDGrid Apps.

1. Docking pre-process softwareCombimark

2. Docking software 1) Dock UCSF 2) gsDock SIMM3. CDB build and maintain S/W

Combilib4. AutoDock5. AutoGrid6. Visualisation 7. Security-related tools

Fixed CDB

start

InputFile

PreprocessDock Drug-like

Analysis

New CDB

Experiment

end

CDBGen.

CDBPara.

DDGrid Resources

Chemical Databases (CDB) Each ligand record in a chemical database represents the 3D structural information of

a compound. The numbers of compounds in each CDB can be in the order of tens of thousands and the database size be anywhere from tens of megabytes to gigabytes and

even terabytes.

1. static databasespurchased from commercial chemical company.

Available Chemical Directory (ACD) Chinese natural product database (CNPD)

SPECS databasechemical ADME/T database, etc.

2. dynamic databasesmade by user own, and deployed automatically.

Deployed commercial CDB (appr.700,000)

Name of Database Description

Specs      Provides about 230,000 compounds

CMC-3D Provides 3D models and important biochemical properties (including drug class, logP, and pKa values) for over 8,400 pharmaceutical compounds.

ACD-3D Provides 200,000 3D compounds commercial available

NCI-3D 213,000compounds with 2D information from the National Cancer Institute

CNPD Collected 12,000 Chinese natural products with chemical structure

TCMD With 9127 compounds and 3922 herbs

Vendor Num. of Mol. Vendor Num. of Mol.

ACB-Eurochem 98603 Maybridge 53042

Ambinter 533866 Nanosyn 68317

Asinex 293385 National Cancer Institute 223536

ChemBridge 562624 Otava 181195

ChemDiv 361859 Peakdale 9632

ComGenex 38590 Pharmeks 116355

Enamine 533111 PubChem 164031

IBScreen 452728 Ryan Scientific 64205

InterChim 288882 Sigma-Aldrich 49022

KeyOrganics 22294 Specs 307550

Life Chemicals 44762 TimTec 127173

appr. 3,300,000 compounds

CDB example ： CNPD-China Natural Products Database

                                                                               

         

                                                                    

         

                                                                                

        

                                                                      

       

                                                                             

           

                                                                  

           

CDB example ： CNPD

CNPD: The first and only comprehensive source of chemical, structural and bibliographic data on all known natural products in China.

CNPD serves as information sources for chemical, physical and biological properties, literature, they are useful to scientists within the pharmaceutical industry.

CNPD can be searched in flexible ways: structure, sub-structure, name, molecular formula, molecular weight, CAS register number, category, etc.

CNPD: Traditional Chinese Medicine (TCM) applications are pre-indexed in CNPD to provide hints for lead compounds discovery.

CDB example ： CNPD

CDB example ： TCMD

TCMD-Traditional Chinese Medicine Database

TCMD is a bibliographical database of approximately 20,000 records with abstracts of TCM articles. Relevant articles are selected from among 150-200 journals from Mainland China, Taiwan, and Hong Kong (most of them are Chinese); English abstracts are written for the selected articles and other pertinent information is translated into English.

CDB example ： TCMD

DDGrid applications in reality

SIMM carried out anti-SARS and anti-diabetes drug research using the DDGrid

1. Anti-SARS drug research2. Anti-diabetes drug research

Virtual screening from Comprehensive Medicinal Chemistry-3D (CMC-3D) database which contains 7,900 compounds, found that cinanserin have distinct anti-SARS effect

Department of Virology, Bernhard-Nocht-Institute for Tropical Medicine, Germany

Research Department, Cantonal Hospital St Gallen, Switzerland

“Basically your inhibitor turned out to be the best compound we have tested so far! ”

Have applied for domestic patent 03129071.x and PCT patent pi034248

Research on Anti-SARS medicine

Found an anti-

diabetes lead better

than Rosiglitazone.

by

targeting on

PPAR ， through

virtual screening,

optimization design

and synthesis and

biology and

pharmacology

testing

CADD process

10

100

1k

10k

100k

1M

10M

2400000

10,000

500300

14276

800,000

200,000

13814

Research on anti-diabetes medicine

2.4 m

10 t

500

142

76protein testing

400 t

85

composite design

virtual screening

virtual screening

48

synthesis

8cell testing

4animal testing

1comprehensive evaluation

48 KD<1 M

22 KD<0.1MKD<100M

protein testing

Research on anti-diabetes medicine

manually screening

New anti-diabetes drug

Current Progress

1. Applied for patent 200410016460.X ， and PCT patent

2. Security testing and pre-clinic research

What does the DDGrid provide ？

1 、 Drug Design Collaboration PlatformLarge-scale Virtual Screening platformsharing large CDB

2 、 Computational Resources SharingSIMM/SSC/HKU/Mol. Ltd/SJTU/DUT

3 、 Data Resources Sharingpre-deployed commercial CDB (ACD/CNPD …)sharing self-made CDB

4 、 Medicinal chemistry text and structure search5 、 Customization and Extension

Collaboration

Selected Users of DDGrid

DDGrid Demo

http://www.ddgrid.ac.cn

Demo

Demo

Demo

Demo

Demo

Demo

Thank you ！

Q&A