The document discusses the importance and challenges of sourcing high-quality online data resources for computational toxicology. It highlights data quality issues in computational models, the variety of available data sources, and includes insights on drug-induced liver injury (DILI) and the role of crowdsourcing in data curation. Additionally, it presents an overview of various databases and emphasizes the future development of an open pharmacological space for better data access and collaboration.

1. Sourcing High-Quality Online Data Resources for Computational Toxicology Antony Williams Bio-IT World, Current Methods for Computational Toxicology and Chemogenomics

2. The Community Depends On Us “ We don’t want another Love Canal!” “ What we know about PCBs should warn us all!” The public is “suspicious” of pharma… “ Chemicals are dangerous”

3. Comp Tox Models Depend on DATA Models for Computational Toxicology depend on the quality of the training set There are multiple issues with data quality including: Experimental The validity of the method, Reproducibility, Sample quality, Data capture, Transcription of values Computational Accurately representing the data – correct units, annotations, quality flags, attribution, are the structures correct?

4. Nothing but the Facts Jean-Claude Bradley, Drexel University “ There are no facts, only measurements embedded within assumptions”

5. Open Notebook Science UsefulChem Blog: http://tinyurl.com/48dyujh

6. Aqueous Solubility of ECGC Epigallocatechin gallate solubility in water

7. Melting Point of DMT

8. Content is King and Quality Costs Chemistry “content” is big money Patent searching Structures and properties Drug databases Literature databases Chemical Abstracts Service (CAS), the “Gold Standard” in Chemistry related information 101 years of content $260 million revenue (2006) >50 million substances >60 million sequences

9. Where can we find data online?

10. Where is chemistry online? Encyclopedic articles (Wikipedia) Chemical vendor databases Metabolic pathway databases Property databases Patents with chemical structures Drug Discovery data Scientific publications Compound aggregators Blogs/Wikis and Open Notebook Science

11. Lots of “Public Compound” Databases PubChem Drugbank ChEBI/ChEMBL KEGG LipidMAPs ChemIDPlus eMolecules ZINC Lots of chemical vendors ChemSpider

12. Toxicology Data

14. Chemistry on the Internet ChemSpider “links” chemistry on the internet Almost 25 million compounds, 400 data sources Allows community deposition, curation, annotation Integrating properties, publications, patents, media Text, structure and substructure searching

15. www.chemspider.com

16. Search for a Chemical

17. Available Information… Linked to vendors, safety data, toxicity, metabolism

18. We Have Delivered the Vision “ Build a Structure Centric Community to Serve Chemists” Integrate chemical structure data on the web Create a “structure-based hub” to information, data and algorithmic predictions Let chemists contribute their own data Allow the community to curate/correct data

19. Dialects describing chemicals

20. What is the Structure of Vitamin K?

21. MeSH A lipid cofactor that is required for normal blood clotting. Several forms of vitamin K have been identified: VITAMIN K 1 (phytomenadione) derived from plants , VITAMIN K 2 (menaquinone) from bacteria, and synthetic naphthoquinone provitamins, VITAMIN K 3 (menadione). Vitamin K 3 provitamins, after being alkylated in vivo, exhibit the antifibrinolytic activity of vitamin K. Green leafy vegetables, liver, cheese, butter, and egg yolk are good sources of vitamin K

22. What is the Structure of Vitamin K1?

23. What is the Structure of Vitamin K1?

24. CAS’s Common Chemistry

25. Wikipedia

28. ChEBI – Manual Curation

31. PubChem

33. “ 2-methyl-3-(3,7,11,15-tetramethyl hexadec-2-enyl)naphthalene-1,4-dione” Variants of systematic names on PubChem 2-methyl-3-[(E,7R,11R)-3,7,11,15-tetramethyl 2-methyl-3-[(E,7S,11R)-3,7,11,15-tetramethyl 2-methyl-3-[(E,7R,11S)-3,7,11,15-tetramethyl 2-methyl-3-[(E,7S,11S)-3,7,11,15-tetramethyl 2-methyl-3-[(E,11S)-3,7,11,15-tetramethyl 2-methyl-3-[(E)-3,7,11,15-tetramethyl 2-methyl-3-(3,7,11,15-tetramethyl 2-methyl-3-[(E)-3,7,11,15-tetramethyl

34. Public Domain Chemistry Databases Our databases are a mess… Non-curated databases are proliferating errors We source and deposit data between databases Original sources of errors hard to determine Curation is time-consuming, challenging and exacting

35. Vancomycin Who will curate? PubChem is not resourced to clean these errors How would you clean such a large dataset?

36. The FDA’s DailyMed

37. Structures on DailyMed

38. Lack of Stereochemisty

39. Incorrect Structures

40. Wow!

41. We want to model DILI… Drug metabolism in the liver can convert some drugs into highly reactive intermediates, This can affect the structure and functions of the liver. Drug-induced liver injury (DILI), is the #1 reason drugs are not approved and withdrawn from market after approval Estimated global annual incidence rate of DILI is 13.9-24.0 per 100,000 inhabitants DILI accounts for an estimated 3-9% of all adverse drug reactions reported to health authorities Herbal components can cause DILI too Thanks to Sean Ekins https://dilin.dcri.duke.edu/for-researchers/info/

42. Initial DILI data – Names and Data Griseofulvin Hycanthone Hydrochlorothiazide Hydrocortisone Hydroxyurea Idarubicin HCl Idoxuridine Imipramine HCl indomethacin isoniazid Isoproterenol HCl Isotretinoin Isoxsuprine HCl Kanamycin Sulfate Ketorolac Tromethamine Ketotifen Labetalol

43. So you want data on drugs??? Sourcing data based on drug names is difficult! Where would you find the “correct chemical structures”? What databases can you trust?

44. Vytorin: Ezetimibe/Simvastatin

45. Vytorin: Ezetimibe/Simvastatin

46. Vytorin: Ezetimibe/Simvastatin

47. Vytorin: Ezetimibe/Simvastatin

48. Vytorin: Ezetimibe/Simvastatin

49. Symbicort: Budesonide + Formoterol

50. Symbicort: Budesonide + Formoterol ChemIDPlus Wikipedia

51. DrugBank: Search Symbicort…

52. Symbicort: Budesonide + Formoterol PubChem 8 structures called Budesonide. 1 “correct” 6 structures called Formoterol. 1 “correct” Search on “Symbicort” gives 1 structure.

53. Taxol: Paclitaxel 44 structures

54. Taxol: Paclitaxel Bioassay Data

55. Public Domain Chemistry Databases An examination of quality in databases – inter/intra lab comparison of processes for 150 drugs

57. Drug Name Generic Name ChEBI ChemSpider CAS Com. Chem ChemIDPlus DailyMed DrugBank PubChem Wikipedia Spiriva Tiotropium Bromide No Hits  No Hits    4/0  Depakote Valproate semisodium        No Structure Basen Voglibose   No Hits  No Hits  2/1  Symbicort 1) Budesonide       8/1  Symbicort 2) Formoterol WRONG  No Hits    6/1  Vytorin 1) Ezetimibe   No Hits      Vytorin 2) Simvastatin       2/1  Taxol Paclitaxel       44/1  Thalidomid Thalidomide No Hits        Zocor Simvastatin       2/1  Crestor Rosuvastatin   No Hits    2/1 

58. Personal Experiences Highest Quality Resources : DSSTox (EPA), ChEBI (EBI) High Quality Resources : DrugBank, Human Metabolome Database, ChemIDPlus, ChemSpider, KEGG Are there others you use???

59. What can be done to help? “ Crowdsourcing” – gather the support of members of the community to add, annotate and curate data Wikipedia is the domain success story for crowdsourcing. PubChem is an example of “crowdsourced deposition” of chemistry data ChemSpider is an example of “crowdsourced deposition and curation”

60. The Future: Open Source and Data Open source software : descriptors and algorithms QSAR should be cheaper and better! Selectively share your models with collaborators Centralized hosting of models / predictions

61. The Future: Open PHACTS The Open PHACTS project will develop an open access innovation platform, called Open Pharmacological Space (OPS), via a semantic web approach. OPS will be comprised of data, vocabularies and infrastructure needed to accelerate drug-oriented research.

62. Exposing Data for Semantic Web…

63. Coming soon… Book chapter: “ Accessing, Using and Creating Chemical Property Databases For Computational Toxicology Modeling ” Antony J. Williams, Sean Ekins, Ola Spjuth and Egon L. Willighagen

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