Validation of biomarkers for drug screening through ordinal logistic regression.

Since the Comprehensive Proarrhythmia Assay (CiPA) initiation, many studies have suggested various features based on ionic charges, action potentials (AP), or intracellular calcium (Ca) to assess proarrhythmic risk. These features are computed through electrophysiological simulations using experimental datasets as input, therefore changing with the quality of experimental data; however, research to validate the robustness of features for proarrhythmic risk assessment of drugs depending on datasets has not been conducted. This study aims to verify the availability of features commonly used in assessing the cardiac toxicity of drugs through an ordinal logistic regression model and three datasets measured under different experimental environments and with different purposes. We performed drug simulations using the Tomek-Ohara Rudy (ToR-ORD) ventricular myocyte model and computed 12 features comprising six AP features, four Ca features, and two ion charge features, which reflected the effect and characteristics of each data for CiPA 28 drugs. We then compared the classific performances of ordinal logistic regressions according to these 12 features and used datasets to validate which feature is the best for assessing the proarrhythmic risk of drugs at high, intermediate, and low levels. All 12 features helped determine high-risky torsadogenic drugs, regardless of the datasets used in the simulation as input. In the three types of features, AP features were the most reliable for determining the three Torsade de Pointes (TdP) risk standards. Among AP features, AP duration at 50% repolarization (APD) was the best when individually using features per dataset. In contrast, the AP repolarization velocity (dVm/dt) was the best when merging all features computed through three datasets.