Real-time prediction of drug-induced proarrhythmic risk with sex-specific cardiac emulators

Abstract

In silico trials for drug safety assessment require a large number of high-fidelity 3D cardiac electrophysiological simulations to predict drug-induced QT interval prolongation, making the process computationally expensive and time-consuming. These simulations, while necessary to accurately model the complex physiological conditions of the human heart, are often cost-prohibitive when scaled to large populations or diverse conditions. To overcome this challenge, we develop sex-specific emulators for the real-time prediction of QT interval prolongation, with separate models for each sex. Building an extensive dataset from 900 simulations allows us to show the superior sensitivity of 3D models over 0D single-cell models in detecting abnormal electrical propagation in response to drug effects as the risk level increases. The resulting emulators trained on this dataset showed high accuracy level, with an average relative error of 4% compared to simulation results. This enables global sensitivity analysis and the replication of in silico cardiac safety clinical trials with accuracy comparable to that of simulations when validated against in vivo data. With our emulators, we carry out in silico clinical trials in seconds on a standard laptop, drastically reducing computational time compared to traditional high-performance computing methods. This efficiency enables the rapid testing of drugs across multiple concentration ranges without additional computational cost. This approach directly addresses several key challenges faced by the biopharmaceutical industry: optimizing trial designs, accounting for variability in biological assays, and enabling rapid, cost-effective drug safety evaluations. By integrating these emulators into the drug development process, we can enhance the reliability of preclinical assessments, streamline regulatory submissions, and advance the practical application of digital twins in biomedicine. This work represents a significant step toward more efficient and personalized drug development, ultimately benefiting patient safety and accelerating the path to market.