Towards closing gaps in livestock physiologically-based kinetic modeling: Machine learning prediction of total plasma clearance and tissue-plasma partition coefficients, and integration of growth in a physiologically-based kinetic model with a case study in PFOS

Physiologically-based kinetic (PBK) models simulate the absorption, distribution, metabolism, and excretion (ADME) of chemicals in the body. A PBK model consists of multiple compartments, representing various bodily tissues, each involving tissue-specific properties and processes. In this thesis, the focus lies on toxicological PBK modeling for livestock species. A PBK model requires a certain number of input parameters, which are often unavailable for livestock species; alternative approaches to parameter procurement need to be found. Part one of the thesis consists of in silico approaches for predicting clearance and tissue partitioning using machine learning (ML). The ML models were based on molecular information, and anatomical/physiological data of pre-clinical and livestock species, such as body weight or tissue composition. The ML predictions were better than those from legacy approaches such as allometric scaling or mechanistic models. Toxicological assessment in livestock often involves long-term exposure scenarios and/or chemicals with long half-lives. Furthermore, livestock experience rapid growth during the juvenile phase, resulting in changes in anatomy and physiology throughout most of their lifespan, with implications for the ADME. In part two of the thesis, growth curves for different livestock species were estimated for implementation in PBK models. A case study of perfluorooctanesulfonic acid exposure in livestock was then conducted, using the growth-integrated PBK model. Inclusion of growth increased prediction accuracy of concentrations in both plasma and tissues. For long-term exposures, and/or juvenile animals, and/or chemicals with long half lives, growth should be accounted for in livestock PBK models.