MACHINE LEARNING APPLICATIONS IN TOXICOLOGY: CURRENT APPROACHES AND FUTURE PERSPECTIVES

Objective: Toxicology faces challenges with increasing chemicals and complex exposure scenarios. This review examines the historical development and current status of artificial intelligence (AI) and machine learning (ML) applications in toxicology, from QSAR models of the 1960s to today's deep learning algorithms. Advances achieved through AI integration in fields such as drug discovery, toxicokinetics, nanotoxicology, and environmental toxicology, along with toxicological endpoints including cardiovascular toxicity, hepatotoxicity, carcinogenesis, genotoxicity, and neurotoxicity, have been evaluated.Result and Discussion: AI technologies offer significant advantages such as reduction in animal experimentation, rapid pattern detection, and toxicity prediction with minimal experimental data. Integrated analysis of genomic, proteomic, and chemical structure data elucidates toxicity mechanisms at the molecular level, while ML models developed for multiple organs contribute to reducing drug attrition rates. Data quality, model validation, and interpretability remain primary challenges to be overcome. In the future, integration of traditional toxicological methods with modern computational approaches will provide more reliable and efficient results in risk assessments, while regulatory authorities need to develop standards for AI-supported models. Multi-omic data integration, personalized toxicology approaches, and development of new ML algorithms that can illustrate toxicity mechanisms in humans represent promising directions for advancing the toxicology sciences.