In Silico Hazard Assessment of Ototoxicants Through Machine Learning and Computational Systems Biology.

Individuals across their lifespan may experience hearing loss from medications or chemicals, prompting concern about ototoxic environmental exposures. This study applies computational modeling as a screening-level hazard identification and chemical prioritization approach and is not intended to constitute a human health risk assessment or to estimate exposure- or dose-dependent ototoxic risk. We evaluated in silico drug-induced ototoxicity models on 80 environmental chemicals, excluding 4 with known ototoxicity, and analyzed 76 chemicals using fingerprinting, similarity assessment, and machine learning classification. We compared predicted environmental ototoxicants with ototoxic drugs, paired select polychlorinated biphenyls with the antineoplastic drug mitotane, and used PCB 177 as a case study to construct an ototoxicity pathway. A systems biology framework predicted and compared molecular targets of mitotane and PCB 177 to generate a network-level mechanism. The consensus model (accuracy 0.95 test; 0.90 validation) identified 18 of 76 chemicals as potential ototoxicants within acceptable confidence ranges. Mitotane and PCB 177 were both predicted to disrupt thyroid-stimulating hormone receptor signaling, suggesting thyroid-mediated pathways may contribute to auditory harm; additional targets included AhR, transthyretin, and PXR. Findings indicate overlapping mechanisms involving metabolic, cellular, and inflammatory processes. This work shows that integrated computational modeling can support virtual screening and prioritization for chemical and drug ototoxicity risk assessment.