Machine Learning-Based Prediction of Drug-Induced Hepatotoxicity: An OvA-QSTR Approach

Drug-induced hepatotoxicity, also known as drug-inducedliver injury(DILI), is among the possible adverse effects of pharmacotherapy.This clinical condition is accepted as one of the factors leadingto patient mortality and morbidity. The LiverTox database was builtby the National Institute of Diabetes and Digestive and Kidney Diseases(NIDDK) to predict potential liver damage from medications and takeappropriate precautions. The database has classified medicines intoseven risk categories (A, B, C, D, E, E*, and X) to avoid medicine-inducedliver toxicity. The hepatic damage risk decreases from group A togroup E. This study did not include the E* and X classes because theycontained unverified and unknown data groups. Our study aims to predictpotential liver damage of new drug molecules without using experimentalanimals. We predict which of the LiverTox risk category drugs withunknown liver toxicity potential will fall into using our one-vs-allquantitative structure-toxicity relationship (OvA-QSTR) model.Our dataset, consisting of 678 organic drug molecules from differentpharmacological classes, was collected from LiverTox. The OvA-QSTRmodels implemented by Bayesian Network (BayesNet) performed well basedon the selected descriptors, with the precision-recall curve(PRC) areas ranging from 0.718 to 0.869. Our OvA-QSTR models providea reliable premarketing risk evaluation of pharmaceutical-inducedliver damage potential and offer predictions for different risk levelsin DILI.