The paper in Frontiers in Physiology is in a special issue covering Interoperability in Scientific Computing and Health Informatics. If you ever work with action potential models, or indeed any stiff* ODEs, then it is probably worth a read.
The paper aims to:
- Demonstrate Chaste’s capabilities for nice handling of CellML models, and the range of solvers that can be used;
- Show the tolerances/timesteps that are required for (i) stability and (ii) accuracy for each model and solver (and highlight the fact you might get plausible looking nonsense if you only check stability!);
- Present some speed benchmarking on the best ODE solver to use for a wide range of action potential models, in two settings: single-cell (the answer is CVODE); and tissue simulation (the answer is – it depends!). We describe some of the tricks and tips we’ve used to get things as fast as possible (analytic jacobians, lookup tables, choice of compiler etc.).
Comments and feedback are welcome.
*loose definition of ‘stiff’ is ODEs with different timescales at different times/in different equations.