Development of AI-Driven Pharmacokinetic Models for Optimizing Drug Dosing

Pharmacokinetics (PK) is the study of the absorption, distribution, metabolism, and elimination of drugs, and provides the rationale for designing appropriate dosage regimens. Traditional compartmental and physiologically based pharmacokinetic (PBPK) models are often used to understand drug disposition and predict exposure. Nevertheless, multiple factors, including age, body weight, organ function, genetic variations, disease severity, concurrent therapies, and other variables, may impact the relationship above and contribute to inter individual variability, making dose selection challenging. With its powerful pattern recognition and predictive capabilities, artificial intelligence (AI) has been increasingly adopted in pharmaceutical research and development. This review focuses on the principles and applications of PK modelling and analysis using AI. The key topics include data preparation, feature selection, hybrid models, model evaluation, model interpretation, and model-informed precision dosing. Applications of AI in drug discovery and development were discussed, with an emphasis on published studies that employed machine learning (ML) algorithms for voriconazole dosing prediction, sitafloxacin PK/PD modeling , extraction of drug–drug interaction features, and prediction of unbound drug exposure in the brain. The discussion demonstrated that computational approaches offer modelling flexibility to address diverse research questions but require careful consideration of data characteristics, analysis endpoints, and evaluation criteria to ensure utility and reliability. By combining AI with conventional pharmacometric tools, valuable insights can be uncovered from heterogeneous and often sparse data domains to facilitate decision-making in dose selection and patient care. At the same time, proper validation strategies, informative model reporting, uncertainty analysis, and patient population evaluation are needed to guide the clinical implementation of AI-enabled models and analyses. In summary, AI-enabled PK modelling and analysis should be regarded as an auxiliary to conventional pharmacological principles and clinical decision-making.