Toxicity Forecasts: Navigating Data-Driven AI/ML Models: From Theory to Practice
B.V. S. Suneel Kumar, Antoine Moitessier, Nicolas Moitessier · Apple Academic Press eBooks · 2025
Drug toxicity plays a crucial role in the withdrawal of drugs from clinical trials. Predicting drug candidates’ toxicity profiles may be considered as the first step towards drugs safety and efficacy. However, this goal may only be reached if predictive tools and models are available. It is believed that the future of toxicity forecasting lies in computational models such as artificial intelligence (AI) and machine learning (ML). However, developing such computational methods has its fair share of challenges, including limited comprehensive datasets, model interpretability, applicability domain (AOD), model biases, and generalizability. Integrating multi-model data sources is a solution to address these challenges and requires understanding theoretical foundations, selecting appropriate algorithms and datasets, and considering model applicability. By overcoming these hurdles and obtaining proper toxicity profiles, drug safety and efficacy may be greatly enhanced. This chapter will cover some of the fundamental steps and concepts in AI/ ML model development from a practical standpoint: dataset collection, molecular featurization, AI/ML theory, validation, and evaluation metrics. The python codes and Jupyter notebooks that are discussed in this book 210 chapter are available in author’s GitHub profile, which can be accessed at: https://www.w3.org/1999/xlink" xlink:href=" https://github.com/suneelbvs/toxicity-forecasts "> https://github.com/suneelbvs/toxicity-forecasts .