Artificial intelligence for toxicokinetic parameterization in environmental exposure assessment

Abstract

Toxicokinetics (TK) characterizes the absorption, distribution, metabolism, and excretion (ADME) of chemicals in the body and connects exposure with internal dose and potential toxicity. Because in vivo and in vitro TK assays are costly and low-throughput, and demand for emerging chemicals is growing, artificial intelligence and machine learning (AI/ML) offer alternative tools to predict ADME profiles for data-poor chemicals. This review examines recent developments in AI-based TK parameterization with relevance to environmental exposure assessment. First, we outline the essential elements of AI-based TK parameterization paradigms, including data sources, molecular representations, and commonly used ML algorithms. Second, the recent applications of AI/ML models across four ADME processes are introduced. Third, we discuss current limitations and future perspectives, including model reliability, interpretability, data availability for environmental chemicals, and integrating AI-predicted TK parameters into mechanistic models, such as physiologically based pharmacokinetic (PBPK) models and other downstream modeling frameworks. Although many existing studies rely on pharmaceutical or mixed chemical datasets, these approaches could support TK parameterization and exposure assessment for environmental chemicals. Further progress will depend on improved data quality, broader environmental chemical coverage, and more rigorous model validation and uncertainty evaluation.