OF THE OXFORD UNIVERSITY COMPUTING LABORATORY PROGRAMMING RESEARCH GROUP STUDENT CONFERENCE ’ 06 Program

Understanding the human heart is a subject of much interest. Heart problems are a majorcause of death in developed countries, and much effort has been put into elucidating thecauses of heart disease in the hope of developing cures. Many drugs also have adverse sideeffects on the heart, which drug companies would like to avoid.Computer simulations have played an important part in this research [4], however sim-ulations are often very time consuming—some can take days to compute even on highperformance computing resources. There is thus considerable interest in optimisation tech-niques, with a view to making whole-heart simulations tractable. Reliability of heart modelsimulations is also of great concern, particularly considering clinical applications. Simula-tion software should be easily testable (against empirical data) and maintainable, which isoften not the case with extensively hand-optimised software. Automating and verifying anyoptimisations is thus crucial.We present a framework for automatically optimising models of the electrical activity ofsingle myocytes (heart muscle cells). This electrical activity is what causes the heart to beat.The models treat a cell like an electrical circuit, representing various features by ordinarydifferential equations (ODEs).CellML [2] is an XML language designed for describing biological cell models from amathematical modeller’s perspective, and is being developed at the University of Auckland.It gives us an abstract format for such models, and provides effectively a structured mathe-matical description of the ODE system. It is gaining in popularity, with various simulationenvironments being able to import CellML models.From a computer science perspective, CellML looks like a domain specific programminglanguage, and so we are investigating the gains available from exploiting this viewpoint. Wedescribe various static (i.e. compile-time) checks for CellML models (including checking fordimensional consistency), and investigate the possibilities of provably correct optimisations.In particular, we demonstrate that partial evaluation [3] is a promising technique for thispurpose, and that it combines well with a lookup table technique, commonly used in cardiacmodelling, which we have automated [1].