A Deep Learning-Based Scoring Framework for Large-Scale Multi-Donor Cardiotoxicity Screening

Cardiotoxicity remains a major cause of drug attrition and post-market withdrawal, yet the vast majority of environmental chemicals to which humans may be exposed remain uncharacterized for cardiotoxicity risk. Human induced pluripotent stem cell (hiPSC)-based testing has been proposed to address this gap. Here we present an unsupervised deep learning framework for multi-donor cardiotoxicity screening using high-throughput calcium transient recordings from hiPSC-derived cardiomyocytes (hiPSC-CMs). We used data from a library of 1,029 compounds that were tested in hiPSC-CM from five donors in concentration-response. An autoencoder trained exclusively on baseline signals quantified chemical-induced functional perturbations through reconstruction error, bypassing the need for labeled training data while capturing the full spectrum of calcium handling disruptions. Using this framework, we generated effect levels. Aggregation of donor-specific scores revealed substantial inter-individual variability in potential cardiotoxicity, underscoring the value of this approach for population-level risk prediction. We found that microbiocides, dyes, and pesticides to have potential concern, characterized by high toxicity scores and low inter-donor variability. This framework establishes a scalable, human-relevant, and genetically diverse platform for cardiotoxicity surveillance across both pharmacological and environmental chemical spaces, with direct implications for drug and chemical safety evaluation and prioritization for additional studies.