Supervisor

Suitable for

Abstract

Our computational models of biological systems are becoming increasingly complicated. Often it is unclear which wet-lab experiments were used for training/calibration/parameter fitting, which have been used for validation of model behaviour (evaluating how predictive the models are), and which experimental results have never been compared with model predictions. It is also difficult to reproduce published results of model simulations, compare the behaviour of different models, determine a model's suitability or limitations for a particular study, or develop models incrementally in a robust manner.

Fundamental to improving this situation is the description of simulation protocols in a machine-readable format, allowing us to replicate wet-lab experiments in-silico, with any of the possible computational models of the biological system. We are developing a system for this, extending the community standard SED-ML - the Simulation Experiment Description Markup Language - for representing protocols. The resulting ability to compare model behaviours under a range of experimental scenarios is a concept we have termed "Functional Curation". It can provide a more comprehensive view of model behaviour, far beyond that described in traditional publications, greatly assisting users in selecting a suitable model for reuse which will, in turn, result in the development of more relevant and accurate models.

You can see the current Functional Curation system in action and experiment with it for yourself at https://travis.cs.ox.ac.uk/FunctionalCuration/index.html

Several projects are available in this area, depending on the interests and experience of the student. In most cases a knowledge of Python and/or C++ would be useful. Some examples are:

• Improving usability, for example by developing an editor for the protocol language.
• Allowing protocols to be run on SBML models, not just CellML as at present, greatly widening applicability of the technique.
• Extending the use of metadata annotations for interfacing between models and protocols, allowing for reasoning over annotations from ontologies.
• Protocol analysis for performance optimisation, especially the use of parallelisation.
• Allowing protocols to be run on GPGPUs.