2. I mproved V irtual S creening S trategies and E nrichment of F ocused L ibraries in A ctive C ompounds U sing T arget- O riented D atabases I mproved V irtual S creening S trategies and E nrichment of F ocused L ibraries in A ctive C ompounds U sing T arget- O riented D atabases ChemAxon 2005 User Group Meeting May 20th, Budapest, Hungary

3. Our Business To provide global solutions in Knowledge Management- based Drug Discovery Meet the needs of pharmaceutical and biotech companies to integrate information and generate knowledge on drug discovery by providing: Target class databases : GPCR, Ion Channel, Kinase, Protease, Nuclear Receptor… Bio-pharmaceutical relevant databases: ADME/Drug-Drug Interactions, hERG Therapeutic area databases : Cancer, Antipsychotics… Build coherent Knowledge Management platform to structure pharmaceutical industry R&D information

5. Current Product Portfolio AurSCOPE Databases Target Based -GPCR -Ion Channel -hERG Biopharmaceutical Topics -ADME/Drug-Drug Interactions AurQUEST Web-based application for querying databases AurSTORE Storage of propriety and third party data Thesaurus and Glossaries for structured data Analysis Application AurTAG: Functional Annotation module

6. AurSCOPE Statistics PublicationsAnalyzedActivitiesLigands Target Proteins GPCR 16000 patents and publications 500,000106,0002300 ADME/ Drug-Drug Interactions 4000 pub92900 8725 parent compound & metabolites 400 HERG450 pub77501000 Ion Channels 4000 patents and publications 12600034000365

7. AurSCOPE GPCR Exhaustively annotated database of GPCR target/Ligand Activity Data All Major GPCR Targets represented Data collected from 370+ journals Historical data 1950s to current Biological Activity data from -Binding, In vivo, Second Messenger, Isolated Organ & other biological protocols

8. AurSCOPE GPCR Data Base Constitution Out of 370 journals

9. JChem Tools Integration Live demo…

10. Representation & Chemical Space Fingerprints calculation Virtual Screening 2D-Chemical similarity 2D-Pharmacophore Similarity Peptides excluded AurSCOPE GPCR 88 209 Query molecules AurSCOPE GPCR Virtual Screening Standardizer PMapperGenerateMD JKlustor

11. PMapperStandardizer ScreenMD GenerateMD HitStatisticsOptimizeMetrics Hits JKlustor ChemAxon Tools

12. NK1 Receptor Klebe et al. J. Med. Chem. 2004, 47, 5381-5392. Modeling of the receptorModeling of the receptor Virtual screening by dockingVirtual screening by docking

13. 2 3 1 Screening Strategy Klebe set Expert set Aureus set

14. Klebe Query Molecules J. Med. Chem. 2004, 47, 5381-5392

15. 3 1 Screening Strategy Klebe set Expert set Aureus set 2

16. Expert Query Molecules J.C. Beaujouan, Collège de France, INSERM U114

17. 2 1 Screening Strategy Klebe set Expert set Aureus set 3

18. Aureus Query Molecules 637 molecules Less than 5nM Number of clusters = 25 Number of singletons = 7

19. Some Aureus Query Molecules More chemical scaffolds Enriching the chemical diversity of query set

20. Number of active NK-1 hits () vs similarity threshold Number of active NK-1 hits ( 100 nM) vs similarity threshold PFCF Klebe set Expert set Aureus set

21. Activity Repartition (sim. threshold = 0.85 ) High (Activity 100 nM) Medium (100 nM Activity 1000 nM) Low (Activity 1000 nM) Klebe (144) (46) (28)Expert (224) (81) (42)Aureus (409) (59) (20) Higher percentage of active molecules with Aureus set

22. Sim = 0.90 254 138 8 1 CFPF Sim = 0.80 470 449 49 34 CFPF Sim = 0.70 688 812 340 1157 CFPF Activity and number of hits vs used fingerprints NK 1 Not NK 1 Non tested NK 1 as potential hits knowing their biological activity on other targets Ideal similarity thresholds to consider for virtual screening of external databases Tanimoto Similarity Threshold

23. Virtual Screening of External Molecular Databases

24. 2.000.000 2.000.000molecules MolLib: Aureus External Molecular Database

25. MolLib Database

26. MolLib Asinex ChemBridge ChemDiv ChemStar … Asinex ChemBridge ChemDiv ChemStar … "Drug-like" Filtering MolLib Representation & Chemical Space Molecular descriptors & Fingerprints Virtual Screening 1) 2D-Chemical Similarity 2) 2D-Pharmacophoric Similarity "Query" Molecules "Consensus" Molecule Hits Focused Libraries AurQuest (Biological Activity) Virtual Screening Strategy of MolLib

27. 287 molecules 100 nM 287 molecules 100 nM AurQUEST Clustering Jarvis-Patrick Algorithm (JChem) Similarity Threshold : 65% Jan. 2005 12 clusters 1e-7 Application: new "Opioid Receptor Like" ligands

28. AurQuest (Jan. 05), JChem clusters NOP Query Molecules ( 100 nM)

29. Tanimoto Similarity threshold JChemCFJChemPF # hits 0.9130 0.82514 0.7-23 0.6-- Nociceptin ligands: Results 1 PF hit : K i = 45 µM 1 CF hit : K i = 297 nM

30. Conclusion Successful integration of ChemAxons cheminformatics toolkit within Aureus Pharmas knowledge management plateforme. Successful integration of ChemAxons cheminformatics toolkit within Aureus Pharmas knowledge management plateforme. Exploitation of AurSCOPE databases in virtual screening strategies. Exploitation of AurSCOPE databases in virtual screening strategies. Rapid 2D similarity search using ChemAxons fingerprints in combination with Aureus-Pharmas diversity-improved molecular sets. Rapid 2D similarity search using ChemAxons fingerprints in combination with Aureus-Pharmas diversity-improved molecular sets.