The role of machine learning in predictive toxicology: A review of current trends and future perspectives

Abstract

Adverse drug reactions (ADRs) are a major challenge in drug development, contributing to high attrition rates and significant financial losses. Due to species differences and limited scalability, traditional toxicity testing methods, such as in vitro assays and animal studies, often fail to predict human-specific toxicities accurately. The emergence of artificial intelligence (AI) and machine learning (ML) has introduced transformative approaches to predictive toxicology, leveraging large-scale datasets such as omics profiles, chemical properties, and electronic health records (EHRs). These AI-powered models provide early and accurate identification of toxicity risks, reducing reliance on animal testing and improving the efficiency of drug discovery. This review explores the role of AI models in predicting ADRs, emphasizing their ability to integrate diverse datasets and uncover complex toxicity mechanisms. Validation techniques, including cross-validation, external validation, and benchmarking against traditional methods, are discussed to ensure model robustness and generalizability. Furthermore, the ethical implications of AI, its alignment with the 3Rs principle (Replacement, Reduction, and Refinement), and its potential to address regulatory challenges are highlighted. By expediting the identification of safe drug candidates and minimizing late-stage failures, AI models significantly reduce costs and development timelines. However, challenges related to data quality, interpretability, and regulatory integration persist. Addressing these issues will enable AI to fully revolutionize predictive toxicology, ensuring safer and more effective drug development processes.