Prediction of Internal Exposures after Virtual Oral Doses of Disparate Chemicals in Rats and Humans Using Simplified Physiologically Based Pharmacokinetic Models with In Silico-Generated Input Parameters.

Abstract

Toxicological evaluation of industrial chemicals with a broad range of chemical structures, for example, bioactive food components, toxic food-derived compounds, and drugs, usually involves the estimation of human clearance by allometric extrapolation of traditionally determined in vivo rat profiles. Three general methods are used to utilize and expand observed time-dependent plasma concentration data after single oral doses of chemicals: empirical standard noncompartmental analysis, compartmental modeling, and physiologically based pharmacokinetic (PBPK) modeling. Application of the PBPK model for forward dosimetry (from external to internal concentrations) following oral administrations has recently been simplified by using in silico-generated input parameters to evaluate internal exposures in humans without reference to any experimental data. Human PBPK model input parameters for a diverse range of compounds have been successfully estimated by using in silico-generated chemical descriptors and machine learning tools. Key values for the fraction absorbed × intestinal availability, the absorption constant, the volume of systemic circulation, and the hepatic intrinsic clearance can be generated in silico using mathematical equations to estimate values for sets of approximately 30 physicochemical properties or in silico descriptors. After virtual oral dosing of more than 350 compounds, the plasma and liver concentrations generated by PBPK models (1) using traditionally determined input parameters and (2) using input parameters estimated in silico were correlated in rat models and human models. This approach to pharmacokinetic modeling could potentially be applied in the clinical setting and during computational toxicological assessment of the potential risks of a wide range of general chemicals.