1. Webinar: Wed 13 th May 2015 UniChem Jon Chambers and Anne Hersey, ChEMBL group, The European Bioinformatics Institute, part of the European Molecular Biology Laboratory (EMBL-EBI). An Introduction to UniChem: EMBL-EBI’s mapping tool for small molecule database identifiers.

2. UniChem Webinar: 13 th May 2015 What is UniChem ? Basic Use of UniChem (web service and web page). Background … Why was UniChem developed ? What problem does it solve ? Requirements and Features… Schema, Data Normalization, Loading Rules, etc Current Content … Sources, Downloads, Stats, Analyses. Connectivity Search. Q and A

3. UniChem Webinar: 13 th May 2015 What is UniChem ? Basic Use of UniChem (web service and web page). Background … Why was UniChem developed ? What problem does it solve ? Requirements and Features… Schema, Data Normalization, Loading Rules, etc Current Content … Sources, Downloads, Stats, Analyses. Connectivity Search. Q and A

4. A ChEMBL Compound Report Card https://www.ebi.ac.uk/chembl/compound/inspect/CHEMBL12

5. Compound Cross-references on a Compound Report Card… Cross-references to the same molecule in other resources. Automatically maintained via UniChem web services. Other resources can make use of this same functionality.

6. REST Web services.

7. REST web services https://www.ebi.ac.uk/unichem/rest/src_compound_id/CHEMBL12/1

8. https://www.ebi.ac.uk/unichem/

9. UniChem query results. LR = Last Release when Assignment was current. UCI = UniChem Identifier

10. UniChem Webinar: 13 th May 2015 What is UniChem ? Basic Use of UniChem (web service and web page). Background … Why was UniChem developed ? What problem does it solve ? Requirements and Features… Schema, Data Normalization, Loading Rules, etc Current Content … Sources, Downloads, Stats, Analyses. Connectivity Search. Q and A

11. EBI Resources containing small molecule data. ‘CHEMBL12’ ‘49575’ ‘DZP’ ‘ECBD..??’ ‘diazepam’ ‘SCHEMBL21442’ -Many resources, each with very different user-bases. -New resources predicted to be developed/adopted in future. -How can chemistry-centric users make use of all these data ? -Links between resources allow each resource to evolve independently. -But, maintenance is manual/time consuming, and a duplication of effort.

12. Advantages of the UniChem model. - All EBI DBs share the maintenance overhead of creating links to each other. UniChem - All EBI DBs share the benefits of maintained links to external resources. - The ‘mapping service’ could be opened for use by external users.

13. UniChem Webinar: 13 th May 2015 What is UniChem ? Basic Use of UniChem (web service and web page). Background … Why was UniChem developed ? What problem does it solve ? Requirements and Features… Schema, Data Normalization, Loading Rules, etc Current Content … Sources, Downloads, Stats, Analyses. Connectivity Search. Q and A

14. Essential requirements for UniChem. Create cross-referencing of chemical structures and their identifiers between databases. Fast (ie: capable of producing mappings ‘on the fly’ during a web page load, via a web service call.) Low maintenance. Up to date. Archive and track changes to ‘id-to-structure’ assignments over time.

15. Standard InChI used as the normalizing mechanism. InChIs (International Chemical Identifier). Non-proprietary, free. Not a registry system. Designed for printed and electronic data sources. Hashed representation aids ‘private’ querying.

16. InChI (International Chemical Identifier) InChIKey… 27 characters long… MGDTEJBDJOHWYU-UHTGSUKQAC-N [ ‘connectivity block’ aka ‘First InChIKey Hash Block’ (FIKHB) shown in blue ]

17. UniChem Schema UC_STRUCTURE UC_XREF UC_RELEASE UC_SOURCE Entries here are immutable eg: CHEMBL12 1 or 0 UCI -PK STANDARDINCHI STANDARDINCHIKEY UCI -FK -PK SRC_ID -FK-PK SRC_COMPOUND_ID-PK ASSIGNMENT LAST_REL_CURRENT SRC_ID-PK NAME DESCRIPTION CURRENT_RELEASE_U etc SRC_ID-PK RELEASE_U-PK SRC_RELEASE_NUMBER SRC_RELEASE_DATE etc

18. UniChem Tracks Historical Assignments… InChiX cpd123 UniChem will record that in this particular source, the id ‘cpd123’… … was last assigned to InChiX on Release No.1, but is not currently assigned to this structure. … was last assigned to InChiY on Release No.2, but is not currently assigned to this structure. … is currently assigned to InChiZ. Data Release No1 from Source ‘S’: Data Release No2 from Source ‘S’: cpd123 InChiY Data Release No3 from Source ‘S’: (latest) cpd123 InChiZ ie: UniChem keeps a record of current AND obsolete assignments.

19. UniChem deals with ‘Multiple Assignments’… InChiX cpd123 cpd456 cpd789 Multiple ids from a particular source assigned to a single InChI… cpd123 InChiX InChiY InChiZ Single id from a particular source assigned to multiple InChIs… …and…

20. Loading Rules Records are not loaded if…  There is a mis-match between the InChI and the InChIKey…  ie: where the InChIKey calculated by UniChem from the InChI provided by the source does not exactly match the InChIKey provided by the source.  The Standard InChI supplied is greater than 2000 characters long. 20

21. Automated Loading and Release. Common Format Source specific downloaders and parsers Single loader Overall process controlled by crontab (timings optimized for each DB to capture latest releases asap). Weekly release process Productio n Release Incl. Downloads+ Mapping files … etc …

22. UniChem Webinar: 13 th May 2015 What is UniChem ? Basic Use of UniChem (web service and web page). Background … Why was UniChem developed ? What problem does it solve ? Requirements and Features… Schema, Data Normalization, Loading Rules, etc Current Content … Sources, Downloads, Stats, Analyses. Connectivity Search. Q and A

23. Top Level stats.

24. Stats. 24 https://www.ebi.ac.uk/unichem/ucquery/stats

25. Sources.

26. Sources

27. Downloads. ftp://ftp.ebi.ac.uk/pub/databases/chembl/UniChem/

28. Downloads on the UniChem ftp site …

29. Oracle Dumps on the UniChem ftp site … Release number == UDRI

30. Contents of a single Release directory…

31. Downloads on the UniChem ftp site …

32. Whole Source Mapping Downloads

33. Whole Source Mapping Downloads – Files containing all id mappings between two sources.

34. An Example of a Whole source mapping file. From src:'3'To src:'15' SX2SCHEMBL3396223 0DUSCHEMBL6234813 FM9SCHEMBL12263874 HHHSCHEMBL1957930 2DCSCHEMBL1746175 28YSCHEMBL232090 0X5SCHEMBL3515230 PU7SCHEMBL1964201 1LPSCHEMBL111850 ACKSCHEMBL4066485... (8719 records) eg: src3src15.txt [PDBe and SureChEMBL]

35. Analyses. Various analyses run on the current UniChem content, using ‘Structural Identity’ defined in one of 3 ways… FULIK = The Full InChIKey. FIKHB = First InChIKey Hash Block (commonly called 'the connectivity layer' of the InChIKey). SCFIB = Separated Single Components of FIKHB.

36. Structures by Source Numbers of ‘structures’ contributed by each source, and of these, how many are unique to the source…

37. Overlaps between Sources Numbers of ‘structures’ which ‘overlap’ between pairs of sources…

38. UniChem Webinar: 13 th May 2015 What is UniChem ? Basic Use of UniChem (web service and web page). Background … Why was UniChem developed ? What problem does it solve ? Requirements and Features… Schema, Data Normalization, Loading Rules, etc Current Content … Sources, Downloads, Stats, Analyses. Connectivity Search. Q and A

39. UniChem Connectivity Search  An advanced use of UniChem which permits searching across UniChem data sources for molecules with the same molecular skeleton as the query, but which may exist in …  Different stereochemical and isotopic forms  Different salt forms or mixtures Funded by FP7 Capacities Specific Programme, grant agreement no. 284209

40. Connectivity Based Searching in UniChem  Standard UniChem links created only on the basis of identical InChIKeys.  Aim: Create links on the basis of common connectivity (but differing elsewhere; stereochemistry, isotopic composition, etc).  Requirements…  Fast (has to be created dynamically).  Identify ‘relationships’ between molecules (eg: “has same connectivity …and is isotopic variant of”)  Link between cpds with common connectivity within mixtures/salts.  Generic / Flexible / Customizable. Funded by FP7 Capacities Specific Programme, grant agreement no. 284209

41. Alternative views of molecular equivalence.  Sometimes, molecules that many scientists would consider equivalent in the context of their particular field (e.g. pharmacology, docking, etc.), are quite often depicted differently across different resources.  Frequently, these depictions have different Standard InChIs and so cannot be integrated by simply matching on Standard InChIKey.  Examples…

42. Isotopic Differences PubChem CID 71450958 CHEMBL441225 DTQNEFOKTXXQKV-XRLBDJASSA-N DTQNEFOKTXXQKV-HKUYNNGSSA-N CP-99994, an NK1 antagonist… NB: First InChIKey Hash Block (FIKHB) in blue.

43. Example of Stereochemical differences AHOUBRCZNHFOSL-WMLDXEAASA-N AHOUBRCZNHFOSL-YOEHRIQHSA-N Paroxetine in two different sources …. Incorrectly drawn, or Valid stereoisomeric forms ? NB: First InChIKey Hash Block (FIKHB) in blue.

44. PIPZGJSEDRMUAW- VJDCAHTMSA-N Yohimbine HCl (Antagonil in ‘Selleck’) Yohimbine (CHEMBL15245 in ChEMBL) BLGXFZZNTVWLAY- SCYLSFHTSA-N Links between mixtures / salts ?

45. QJVHTELASVOWBE-AGNWQMPPSA-N Amoxicillin Clavulanic acid Co_Amoxiclav InChI=1S/C16H19N3O5S.C8H9NO5/c1-16(2)11(15(23)24)19- 13(22)10(14(19)25-16)18-12(21)9(17)7-3-5-8(20)6-4-7;10-2-1-4- 7(8(12)13)9-5(11)3-6(9)14-4/h3-6,9-11,14,20H,17H2,1- 2H3,(H,18,21)(H,23,24);1,6-7,10H,2-3H2,(H,12,13)/b;4-1-/t9-,10-,11+,14-;6-,7-/m11/s1

46. PIPZGJSEDRMUAW- VJDCAHTMSA-N Yohimbine HCl Yohimbine BLGXFZZNTVWLAY- SCYLSFHTSA-N Links between mixtures / salts ?

47. PIPZGJSEDRMUAW- VJDCAHTMSA-N Yohimbine HCl Yohimbine BLGXFZZNTVWLAY- SCYLSFHTSA-N Links between mixtures / salts ? BLGXFZZNTVWLAY- SCYLSFHTSA-N VEXZGXHMUGYJMC- UHFFFAOYSA-N Hydrochloride Yohimbine …Yes, but parsing of the InChI required first...

48. UniChem Schema UC_STRUCTURE UC_XREF UC_RELEASE UC_SOURCE eg: CHEMBL12 1 or 0 UCI -PK STANDARDINCHI STANDARDINCHIKEY FIKHB UCI -FK -PK SRC_ID -FK-PK SRC_COMPOUND_ID-PK ASSIGNMENT LAST_REL_CURRENT SRC_ID-PK NAME DESCRIPTION CURRENT_RELEASE_U etc SRC_ID-PK RELEASE_U-PK SRC_RELEASE_NUMBER SRC_RELEASE_DATE etc UC_FIKHB_HIERARCHY PARENT CHILD Additions to schema for ‘Connectivity Search’ shown in green

49. BLGXFZZNTVWLAY- SCYLSFHTSA-N Links between combinations of stereoisomers, isotopic variants, in mixtures / salts … Yohimbine (CHEMBL15245 in ChEMBL) Yohimbine HCl PIPZGJSEDRMUAW- VJDCAHTMSA-N …is a component of… XIIDGINYXKOJGX-ZKKXXTDSSA-N Rauwolscine Oxalate Rauwolscine HCl PIPZGJSEDRMUAW-ZKKXXTDSSA-N …is a component of… AND …is stereoisomer of… tritiated Rauwolscine BLGXFZZNTVWLAY-XDGRAVGFSA-N …is isotopic variant of… AND …is stereoisomer of…

51. Refining ‘Connectivity Search’ to show salts and mixtures. Select radio button ‘4’ of Option C.

52. Connectivity Search Results Page.

53. Connectivity Search Web Services

54. Connectivity Search Web service query results https://www.ebi.ac.uk/unichem/rest/cpd_search/CHEMBL15245/1/0/0/4

55. Connectivity Search in ChEMBL

57. Train Online http://www.ebi.ac.uk/training/online/course/unichem-quick-tour-0

58. Acknowledgements ChEMBL John Overington Anne Hersey Anna Gaulton Mark Davies Louisa Bellis George Papadatos Shaun McGlinchey Jon Chambers ChEBI Chris Steinbeck Janna Hastings PDBe Sameer Velankar Atlas Robert Petryszak Training Tom Hancocks Richard Grandison

59. UniChem Webinar: 13 th May 2015 What is UniChem ? Basic Use of UniChem (web service and web page). Background … Why was UniChem developed ? What problem does it solve ? Requirements and Features… Schema, Data Normalization, Loading Rules, etc Current Content … Sources, Downloads, Stats, Analyses. Connectivity Search. Q and A

60. 20 th May - ChEMBL walkthrough 27 th May - Sequence searching (*3pm UK time) 3 rd June – UniProt – accessing protein data programmatically 10 th June – MyChEMBL walkthrough 17 th June - ChEMBL Web Services All webinars @ 4:00pm UK time unless stated For details see: http://www.ebi.ac.uk/training/online/embl- ebi-training-webinar-series-2015 Future webinars:

61. __END__

62. 62 InChI=1S/C10H6N4O2/ c15-9-7-8(13-10 … 37325 37327 Example of multiple ids from a source assigned to a single Standard InChI… mappings generated… ChEMBL -> ChEBIChEBI -> ChEMBL CHEMBL68500 -> 3732537325 -> CHEMBL68500 CHEMBL68500 -> 3732737327 -> CHEMBL68500 Mapping imprecision alloxazine isoalloxazine