1. ChemSpider – A Crowdsourcing Environment for Hosting and Validating Chemistry Resources (and lessons from President Bush) Antony Williams 5th Meeting on U.S. Government Chemical Databases and Open Chemistry August 2011

2. I want to know about “Vincristine”

3. Vincristine: Identifiers and Properties

4. Vincristine: Vendors and Sources

5. Vincristine: Patents

6. Vincristine: Articles

7. Vincristine: RSC Databases

8. Searches: The INTERNET

9. Validated Names for Searching…

10. And InChIs…

11. ChemSpider  The Free Chemical Database  A central hub for chemists to source information  >26 million unique chemical records  Aggregated from >400 data sources  Chemicals, spectra, CIF files, movies, images, podcasts, links to patents, publications, predictions  A central hub for chemists to deposit & curate data

12. Essential aspects of ChemSpider  ChemSpider is a BIG database..and growing  Our focus has increasingly become QUALITY over quantity  Data curation and validation is our strength – crowdsourcing is contributing, more is required  Validated data has enabled linking of the internet

13. There are NO errors in ChemSpider

15. “All That Glisters is Not Gold” What is the structure of Discodermolide?

16. How to distinguish…who’s wrong?

17. Neither is wrong

18. Data Curation…long torturous task  Data curation – JUST structure-name validation is a long, torturous, iterative task.  How about validating “data” – PhysChem data such as logP data, boiling points, melting points (J.C.Bradley’s talk), spectra

19. Hand on my heart….

20. Hand on my heart  No offence meant by what follows! We ALL have quality issues!

21. PHYSPROP Database The freely downloadable database under the EPI Suite prediction software Very Basic filters suggest data quality issues

22. The Stereochemistry challenge. 12500 chemicals with “missed” stereo

23. NIST Webbook

24. EPA’s DailyMed

27. PubChem

28. Linking

33. Patents

35. WYSIWYG compounds

37. Data Curation…long torturous task  Data curation – JUST structure-name validation is a long, torturous, iterative task.  How about validating “data” – PhysChem data such as logP data, boiling points, melting points (J.C.Bradley’s talk), spectra  The crowd in crowdsourcing is …generally small  Which of the large databases are doing careful curation. How can we share the workload? Hmm..

38.  Consider searching each of these chemical databases by chemical name (systematic name, trade name or synonym). Please mark each online resource according to how much you generally trust the results.

40. Drug NameGeneric Name ChEBIChemSpider CAS Com. ChemChemIDPlusDailyMedDrugBankPubChemWikipedia Spiriva Tiotropium Bromide No Hits   4/0 Depakote Valproate semisodium No Structure BasenVoglibose No Hits  2/1 Symbicort1) Budesonide  8/1 Symbicort2) Formoterol WRONG  No Hits  6/1  Vytorin1) Ezetimibe No Hits Vytorin2) Simvastatin  2/1  TaxolPaclitaxel  44/1 ThalidomidThalidomide No Hits ZocorSimvastatin  2/1 CrestorRosuvastatin No Hits  2/1

41. Who does the Curation?

42. ChemSpider can “do it” for us  ChemSpider has built a curation interface used by the community and ourselves for curating.  All curation activities are available for review, online immediately, iteratively checked.  Curators have different abilities based on their profile: There are only a few “Master Curators”.  Can we “share” the curation workload?

43. Proof of Concept Data Curation Sharing

44. Identifier Dictionaries  Reciprocal curation processes…share curation with each other.  If a database has a compound already then use InChiKeys to match “suggested” validation against the compound.  A series of “added” and “removed” synonyms against InChIKeys for matching.  Who will participate???

45. Proof of Concept Data Curation Sharing

46. Lessons Learned : Big vs Good!

47. 15 compounds called Yohimbine 54 Skeletons for Yohimbine

48. Aggegators suffer dilution…

49. User Understanding of Data  Users searching “Yohimbine” expect to find it…not labeled versions of it, not ambiguous stereochemistries, not partial stereochemistries.  Data “aggregation” into a meaningful form is a major challenge. e.g. Assays for radiolabeled compounds linked to actual drugs.  Data curation efforts such as ChEMBL are essential!

50. SciMobileApps.com

51. SciDBs.com (Coming soon)

52.  Open PHACTS : partnership between European Community and EFPIA  Freely accessible for knowledge discovery and verification.  Data on small molecules  Pharmacological profiles  Pharmacokinetics  ADMET data  Biological targets and pathways  Proprietary and public data sources.

53. Standardization and Quality  Our initial approaches to standardization were imperfect. We are revisiting to support OpenPHACTS.  Highly dependent on InChI and not enough standardization prior to InChI generation.  InChI is excellent and acknowledged imperfect. Way better than SMILES for linking the internet!

54. Conclusions  ChemSpider is one of many important chemistry resources on the internet  We have assumed an important role of curating and validating data – specifically name-structure dictionaries are of high importance but data validation is also key  We are a part of the federation of internet databases serving chemistry. MORE collaboration can serve us all better…how?

55. A Plea to Gov’t DBs…  Please improve gov’t DB communications

56. A Plea to Gov’t DBs…  Please improve gov’t DB communications  Please buddy up and get closer together

57. A Plea to Gov’t DBs…  Please improve gov’t DB communications  Please buddy up and get closer together  Get into deep conversations

58. Acknowledgments  Our development team – headed by THAT man..  Many in this room: InChI, PubChem, DssTOX, FDA, ChEBI/ChEMBL, SureChem, many more  Curators – special gratitude to Barrie Walker!  Software providers – OpenEye, ChemDoodle, ACD/Labs, GGA Software, Open Source (Jmol, JSpecView, OpenBabel)

59. Thank you Email: williamsa@rsc.org Twitter: ChemConnector Personal Blog: www.chemconnector.com SLIDES: www.slideshare.net/AntonyWilliams