ti met may10

Download Ti met may10

Post on 25-May-2015

369 views

Category:

Documents

2 download

Embed Size (px)

TRANSCRIPT

  • 1. The Systems Biology Software Infrastructure TiMet WorkshopMay 7 th2010, Edinburgh Richard Adams www.sbsi.ed.ac.uk http://sourceforge.net/projects/sbsi/

2. SBSI - Overall objective A new infrastructure to streamline the connection between data, models, and analysis, allowing the updating of large scale data, modelsand analytic tools with greatly reduced overhead 3. SBSI Contributors Core developers EPCC Test Models andEvaluationProject management Circadian clock modellers Stephen Gilmore PI Nikos Tsorman Neil Hanlon Galina Lebedeva Alexey Goltsov Azusa Yamaguchi Kevin StratfordPeople previouslyinvolved with SBSI Shakir Ali Anatoly Sorokin Treenut Saithong Stuart Moodie Ozgur Akman Igor Goryanin Biopepa integration Adam Duguid Richard Adams Requirements & NumericsAndrew Millar Carl Troein 4. Graphical Notation Network Inference Process Algebras Model analysis Existing knowledge High-resolution data High-throughput data New knowledge Static models Kinetic models Systems Biology Software Infrastructure Kinetic Parameter Facility Circadian clock RNA metabolism Interferon signalling Systems Biology Research, CSBE view ERB-b signalling 5. Initial use case : Parameter Estimation Problem

  • Building predictive models challenging problem in Systems Biology
  • Parameter estimation critical stage in model development
  • Multiple data sets needed for model calibration
  • Optimization of large scale models computationally challenging
  • Circadian clock modelling project requires model optimization.

6. SBSI Numerics optimization

  • SBML model,
  • Parameter constraints,
  • Experimental Data files
  • Configuration file

Model.cpp, Datafiles, Parameter constraints SBML->C++ conversion

  • Best parameters
  • Cost function behaviour
  • Time course with best parameters

Eddie (ECDF) Output Using command line client Run on HPCretrieve results Input 7. Integration of other CSBE projects BioPepa Outline of SBSI design External model & experimental d ata sources BioModels

  • SBSI
  • Dispatcher
  • (Task Manager )
  • Compile C codes
  • Submit jobs to HPC
  • Retrieve results
  • Provide job status

SBSI Numerics core

  • SBSI Visual
  • Desktop application
  • Upload and edit SBML models
  • Run simulations
  • Configure and run optimisations
  • Interact with external repositories
  • Visualisation of data and results

Eddie (ECDF) SBSI Numerics SBSI Numerics SBSI servers SBSI Numerics SBSI- complete system 8. Integration of otherCSBE projects BioPepa EPE External model & experimental d ata sources

  • SBSI Visual
  • Desktop application
  • Upload and edit SBML models
  • Run simulations
  • Configure and run optimisations
  • Interact with external repositories
  • Visualisation of data and results

SBSI Numerics SBSI- local mode 9.

  • SBSI
  • Dispatcher
  • (Task Manager )
  • Compile C codes
  • Submit jobs to HPC
  • Retrieve results
  • Provide job status

SBSI Numerics CellDesigner Eddie (ECDF) SBSI Numerics SBSI Numerics SBSI servers SBSI Numerics A plugin for CellDesignerCellDesigner SBSI plugin 10. Nactem CellDesigner Dunnart InSilicoIDE SBSI PathText Kleio Panther Pathways database autolayouts visualizes annotates Provides SBML models Optimizes? Sabio- RK database Kineticparameters Copasi Ananiadou/Tsujii/Kemper Mi (SRI) Funahashi/Ghosh Nomura (Osaka) updates EHMN Goryanin (Edinburgh) Boyd (Melbourne) 4-6 July Manchester, 8-9 October Edinburgh (ICSB), OIST early March 2011 Existing organisations/interactions Planned collaborations Gilmore (Edinburgh) GARUDA partners Proposed collaborations 11. Multiple Cost Function 12. Optimizing Circadian Clock models with experimental data BIOMD055: Extension of a genetic network model by iterative experimentation and mathematical analysis. byJ. C. W. Locke, M. M. Southern, L. Kozma-Bognar, V. Hibberd, P. E. Brown, M. S. Turner, A. J. Millar (2005b). Molecular Systems Biology. 1:13 The model has 57 parameters and 13 states( equations). Fitting data is 2 of those states obtained by experiment. Using BG/L 128 nodes, it finished at 63140thgeneration bynon-improvement criteria. The run time is 46 hours.Multiple Cost Function is used up to 6740 generation,after 6740th, only X2Cost is applied 13. Release code base on Sourceforge Establish SBSI Numerics on Hector Provide access to SBSI through CellDesigner Develop user base /community Publish! SBSI goals 2010 14. In the workspace youcanstore models, data, objective functions and results Editor view allows access to files Data visualization panel Step 1 create a newSBSI project Runningparameter optimisations 15. Runningparameter optimisations

  • Step 2 choose models,
  • data and algorithm type
  • multiple datasets can be selected

16. Step 3: choose parameters, constraints and initial values Runningparameter optimisations 17. Runningparameter optimisations Step 4: configure optimization algorithm 18. Step 5:Compare simulation using best parameters,with experimental data.