Informatics in the Manchester Centre for Integrative Systems Biology

1. Informaticsin theManchester Centre for Integrative Systems Biology Daniel Jameson, Neil Swainston Manchester Centre for Integrative Systems Biology SysMO-DB Workshop – Connecting Models and Data, Berlin 23 November 2009

2. The MCISB • Currently employs 9.5 multidisciplinary people • All share same office, lab • Pioneer the development of new experimental and computational technologies in systems biology • Develop an annotated, kinetic model of yeast metabolism

3. Goals of the MCISB • Follow an integrative approach:

4. Goals of the MCISB • Follow an iterative approach:

5. Definition of the problem • Experimentalists generate data • Modellers require data • How do we pass data from the experimentalist to the modeller? • Traditional method • Experimentalist analyses data, produces spreadsheet • Experimentalists e-mails spreadsheet to modeller • Modeller cuts-and-pastes data into modelling tool • Do the experimentalist and the modeller speak the same language?

6. Informatics challenges • How do we map experimental data to models? • How do we know what data applies to what molecule or reaction? • How do we identify molecules or reactions? • (Same problem in merging models) • Use names…?

7. Computers don’t like names …because they are non-unique / ambiguous / imprecise / etc.

8. Biochemists like names a little too much… (3R,4R,5S,6S)-6-(hydroxymethyl) oxane-2,3,4,5-tetrol Glucose Glc Anhydrous dextrose Cerelose 2001 Traubenzucker Staleydex 95M

9. Solution • Utilise unique, public identifiers for identifying molecules • Don’t re-invent your own… • Use ChEBI terms to uniquely identify metabolites • Use UniProt terms to uniquely identify enzyme

11. Solution • Further advantage: • Using links into existing databases (ChEBI, UniProt) provide additional information immediately • Chemical formulae, structures • Protein sequences, phosphorlyation sites, SNPs • Use unique, public IDs

12. But names are still important • Names are for humans (human-ish) • Unique ids (e-mail addresses, bank account numbers) are for computers (geek-ish) • BOTH are needed

13. But names are still important

14. Models • Useful to have a standard to allow models to be shared / re-used • Use SBML • Very well developed / supported • Tool set increasing all the time • Identifying metabolites / proteins in models? • Use MIRIAM standards • http://www.ebi.ac.uk/miriam/ • Allows unique, public IDs to be embedded into SBML as annotations (along with human-readable names)

15. Models • Genome-scale SBML model of yeast metabolism • Annotated model • All >2000 molecules have unique database references • MIRIAM standards have been followed • Should be entirely unambiguous for third party users • Should be usable in third party tools • Should allow data to be imported “easily”

17. SBML annotation <species id=”glc" name="D-Glucose"> <annotation> <rdf:li rdf:resource="urn:miriam:obo.chebi:CHEBI:17634"/> </annotation> </species>

18. Solution on the experimental side • Ensure that unique identifiers are captured and associated with data at the time of the experiment • BUT… this is all a bit geek-ish for biologists • So… generate intuitive tools to do this by stealth

19. KineticsWizard

20. Project overview Enzyme kinetics Quantitative proteomics Quantitative metabolomics PRIDE XML MeMo MeMo-RK SABIO-RK Web service Web service Web service Web service Variables (metabolite, protein concentrations) Parameters (KM, Kcat) SBML Model

24. CellDesigner plugins …eventually

25. But… • …MCISB has to manage “only” three types of experiment • Proteomics, metabolomics, enzyme kinetics • Informatics team share office with experimentalists and modellers • We’ve been doing this for years… • Lots of time, lots of people, lots of resource • Infrastructure development is part of our remit

26. And… • …SYSMO projects are far more diverse • Informatics team separated from experimentalists, who are separated from modellers • Less informatics resource • Heavyweight approach of MCISB (bespoke tools for each experiment) probably not applicable

27. So… • …lightweight approach may be more suitable • Store only secondary data necessary for modelling • Not raw data • Daniel…

28. Einfach Klasse!

29. Modelling infrastructure

30. Taverna http://taverna.sourceforge.net

31. Taverna

32. Modelling life-cycle workflows

33. Model construction Input: list of ORFs 1. Get reaction info 2. Create compartments Get annotations 3. Create species 4. Create reactions Output: SBML file

34. Model construction

35. Model parameterisation • Data requirements • SBML model • Starting concentrations for enzymes and source metabolites • Key results database • Enzyme kinetics • SABIO-RK database web service

36. SABIO-RK web service

37. Model parameterisation

38. Model calibration • Data requirements • Parameterised SBML model • Experimental data • Metabolite concentrations from key results database • Calibration by COPASI web service

39. COPASI web service Design and Architecture of Web Services for Simulation of Biochemical Systems. Dada JO, Mendes P. Data Integration in the Life Sciences, Manchester, UK (2009).

40. Model calibration

41. Model simulation • Using COPASI web service

42. Conclusion • Integrating experimental data with models is “easy” and can be automated • If we adopt some standards • Data can be shared “easily” between groups • If we all adopt some standards • Lightweight approach more achievable • Key Results Database

43. Thanks…

44. Informaticsin theManchester Centre for Integrative Systems Biology Daniel Jameson, Neil Swainston Manchester Centre for Integrative Systems Biology SysMO-DB Workshop – Connecting Models and Data, Berlin 23 November 2009