Derivative-based Inference for Cell and Channel Electrophysiology Models

This repository contains the code for the CINC 2022 paper "Derivative-based Inference for Cell and Channel Electrophysiology Models".

Requirements

• Python 3.7 or newer.
• Myokit (github version, 2022-09-12)
• PINTS (github version, 2022-09-12)

Scripts

Models, protocols, shared code:

• Model and protocol files are stored in resources.
• Python modules used in fitting and stored in methods.
• Results are written to the results directory.

Benchmarking simulations:

• cost.py runs simulations from randomly sampled starting points and measures run times for simulations with and without derivative calculation.

Fitting:

• fit.py performs a fit; see python fit.py --help for details.
• count.py shows the number of fits performed for each method and test case.
• best.py shows the lowest error returned for each method and test case.
• time.py shows the mean time taken for runs getting within 5% of the best result, for each method and test case.

Figures

Rendered figures are stored in figures. All figures can be generated in poster format by adding the command line argument poster.

• m1-cases.py generates a figure showing the results of a simulation with each test case (not included in paper due to space constraints).
• m2-eval-cost.py generates a figure comparing the cost per evaluation with and without derivative calculations (requires cost.py to have been run).
• m3-opt-ikr.py generates a figure showing the fitting error as a function of the number of evaluations and as a function of the run time, for the IKr case.
• m4-opt-ap.py generates a figure showing the fitting error as a function of the number of evaluations and as a function of the run time, for the AP case.
• m5-robustness.py generates a figure comparing "robustness" of fitting methods, for both cases.