Abstract Fri055: Machine Learning-Based Drug Screening Using Human iPS Cell-Derived Cardiac Tissues Identified A Novel Drug for Doxorubicin-Induced Cardiomyopathy

Background: Cardiotoxicity is a life-threatening side effect of anti-cancer drugs like Doxorubicin (DOX). However, no effective treatments currently exist to mitigate drug-induced cardiac damage. Research Questions: Can machine learning-based evaluation of sarcomere structures in human induced pluripotent stem cell (hiPSC)-derived cardiomyocytes and contraction measurement in 3D functional cardiac tissues be applied for high-throughput screening for novel therapies? Goals: To discover new therapeutics to prevent DOX-induced cardiac dysfunction. Methods: We established a high-throughput screening assay integrating machine learning-based evaluation of cardiac damage in hiPSC-derived mature ventricular cardiomyocytes and functional evaluation of 3D mature cardiac tissues composed of hiPSC-ventricular cardiomyocytes and epicardial cells. Results: DOX exposure caused sarcomere disorganization in 2D-cultured mature ventricular cardiomyocytes and contractile impairment in 3D cardiac tissues. We developed a supervised machine learning system to quantify sarcomere disorganization in 2D cardiomyocytes, enabling high-throughput screening and leading to the discovery of a novel drug that protects against DOX-induced sarcomere damage. The cardioprotective effect was validated in 3D mature cardiac tissues, where the drug improved contractile function under DOX exposure. Furthermore, in a DOX-induced cardiomyopathy mouse model, it also enhanced cardiac function and significantly increased survival. Conclusion: This study highlights the potential of combining in vitro disease modeling with machine learning-based analysis for drug discovery. Our findings suggest that this novel drug could be an effective treatment for DOX-induced cardiomyopathy.