Drug induced TdP risks classification assay using electro-mechanical models of human ventricle based on CiPA...

Predicting drug-induced cardiac toxicity is critical for drug safety assays, especially when evaluating risks of inducing Torsade de Pointes (TdP). This study proposes an integrated electromechanical model of myocytes for TdP risk assessment, extending the CiPA framework. Human electrophysiological models of CiPAORdv1.0, ORD, and ToR were integrated with a Land mechanical ventricle model. Twenty-seven parameters were observed, including the net current ( $\mathrm{qNet}$ ), inward current ( $\mathrm{qInward}$ ), action potential profile, intracellular calcium profile, and tension profile. We used ordinal logistic regression with 12 drugs as training dataset and validated using unseen data from the remaining 16 drugs following the protocol from the FDA. We observed that the electromechanical model improved the TdP risk classification in most of the parameters derived from the action potential, Calcium Transient ( ${\mathrm{Ca}}_i$ ), and tension profile. The CiPAORdv1.0 + Land not only preserved $\mathrm{qNet}$ performance but improved the classification performance using ${\mathrm{APD}}_{50}$ , ${\mathrm{APD}}_{90}$ , ${\mathrm{CaD}}_{90}$ , ${\mathrm{Ca}}_{\mathrm{tri}}$ , ${\mathrm{ti}}_{\mathrm{tri}}$ , and $\mathrm{EMW}$ . On the other hand, ToR + Land model improved the parameter of ${\mathrm{CaTD}}_{50}$ and ${\mathrm{CaD}}_{50,tp}$ . The ORD + Land model showed some improvement in $V_{\max }$ , and some well for other parameters including ${\mathrm{CaTD}}_{90}$ , ${\mathrm{CaD}}_{90,tp}$ , ${\mathrm{Ca}}_{\mathrm{tri}}$ and ${\mathrm{ti}}_{\mathrm{tri}}$ . This work highlights the advantage of an electromechanical ventricular model for assessing drug-induced TdP risk compared to an electrophysiological model. Overall, the coupled models improved the TdP classification performance based on APD and calcium as well as tension profile. Further optimization of the models and inclusion of more drugs in training the models can improve interpretability and predictive accuracy for TdP risk assessment.

The online version contains supplementary material available at 10.1007/s43188-025-00320-0.