Assessing Ionic Current Blockades and Electromechanical Biomarkers' Interrelations Through a Novel Multi-Channel Causal Variational Autoencoder

Drug-induced Torsade de pointes (TdP) is a critical arrhythmia that can lead to sudden cardiac death.Besides ionic current blockades, in-silico electrophysiological and mechanical biomarkers can provide mechanistic proarrhythmic information for TdP-risk assessment, and specific torsadogenic indices have been developed for that purpose, yet, determining the causal relationships between these variables is challenging.This study aims to employ a novel Multi-Channel Causal Variational Autoencoder (MC 2 VAE), to identify causal relationships between ionic current blockades, electrophysiological biomarkers, and torsadogenic indices, considered as three distinct sources of information (channels) for drug-induced TdP risk.Our approach interestingly suggests the existence of latent causal relationships between the considered channels, allowing for a better reconstruction of all observed features.Further, MC 2 VAE can quantify the strengths of the identified causal relationships, opening up a viable avenue for actionable interventions on the established causal graph.Finally, we consider the downstream task of drug classification for TdP risk on the latent channels and show a clear improvement in classification performances when combining the three considered channels.Overall, our results provide a strong rationale for causally combining multi-channels biomarkers in TdP-risk characterization.