Digital Twin for the Win: Personalized Cardiac Electrophysiology.

Abstract

Individual variability drives differences in cardiovascular disease manifestation, therapeutic response, and rare phenotypes. Traditional preclinical models minimize variability, limiting their ability to capture population heterogeneity and contributing to translational gaps and adverse drug reactions. Here, we present a computational, experimental, and machine learning framework for generating digital twins of induced pluripotent stem cell-derived cardiomyocytes (iPSC-CMs) that reproduce cell-specific electrophysiology from a single optimized voltage-clamp protocol. A synthetic population of iPSC-CMs was generated by varying 52 parameters across six ionic currents and used to train a neural network to infer parameters from current responses. Applied to experimental data, this approach produced digital twins that captured action potential diversity across temperatures and morphologies. Importantly, baseline variability in ionic current dynamics not only explained heterogeneity in spontaneous action potential waveforms but also predicted differential drug sensitivity, supporting digital twin application in cardiac safety pharmacology and precision medicine.