Machine learning to predict food effects during drug development: a comprehensive review

Drug absorption can be altered due to the consumption of food, impacting the efficacy and safety of the drug administered, and predicting food effects (FE) can be quite complex. Traditional methods, including in vitro and in vivo models, fail to predict the full range of food-drug interactions owing to the biological variability of the gastrointestinal system. This review evaluates the predictive ability and accuracy of machine learning (ML) in predicting FE in comparison to conventional methods. We consider how ML models use food dataset information and assist in enhancing the formulation and dosing of the drugs. We discussed recent trends in FE prediction, its mechanisms, and effects on drug bioavailability. Supervised and unsupervised learning, as well as reinforcement learning, are analyzed in the context of absorption, distribution, metabolism, and elimination (ADME) forecasting and drug development. ML is certainly useful in addressing the issues posed by traditional methods; however, challenges about data quality, model generalizability, and integration into the drug development process are obstacles that must be overcome. This review explains how other emerging technologies, for example, PBPK modeling, can be combined with ML to enhance its prospects in the field of drug development. We examined prospects of deep learning, explainable artificial intelligence (AI), and ethical and legal aspects of applying ML in pharmacokinetics, as well as the interdisciplinary approaches that are required to improve patient care outcomes.