This folder contains source code to reproduce the figures for "Prediction of arrhythmia susceptibility through mathematical modeling and machine learning" published in PNAS.
MATLAB - version 2019b or higher; 2019b was used to run all simulations.
Figures can be recreated by running the "FigureX.m" files. Functions used within each m-file are within the functions folder and have been extensively commented for easy usability.
Factors that can promote ventricular arrhythmias include congenital mutations, certain therapeutics, and structural cardiac remodeling. Additional aspects that may contribute include electrolyte imbalances and sex differences in cardiac electrophysiology. Despite our understanding of the importance of these variables, it remains extremely difficult to predict which specific individuals within a population will be especially susceptible to ventricular arrhythmias. This study presents a novel computational framework that combines supervised machine learning algorithms with population-based cardiac myocyte mathematical modeling. Using this approach, we identify specific electrophysiological signatures that classify arrhythmia response to three individual triggers. Our new predictors significantly outperform the standard myocyte metric, action potential duration (APD). We also find that these new features provide insight into the complex mechanisms that differentiate a susceptible cell from a resistant one. Overall, our pipeline improves on current methods and suggests a proof of concept at the cellular level that can be translated to the clinical level.
Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.