Metabolic Activations and Pharmacokinetics of Drugs Using Data-driven Modeling

Pharmacological and toxicological assessments of a wide array of chemicals typically involve estimating human clearance through allometric extrapolation of in vivo animal data, using empirical compartmental and physiologically based pharmacokinetic (PBPK) models. Parameters for pharmacokinetic absorption, distribution, and metabolic clearance can be calculated and applied to models that replicate observed plasma drug concentration–time profiles. In patients with genetically impaired cytochrome P450 (P450) enzymes who are prescribed only certain drugs, moderate plasma exposure should be noted as a precaution on drug labels, as genetic variants can cause changes in blood concentrations similar to those caused by drug–drug interactions. For new approach methodologies that do not rely on experimental data, human PBPK model input parameters for various compounds have been effectively estimated using in silico-generated chemical descriptors and machine learning tools to assess internal exposure in humans. Many drug oxidations are facilitated by species-dependent and polymorphic P450s and flavin-containing monooxygenases (FMOs). For instance, in rats, the main oxidation product of thalidomide was deactivated 5′-hydroxythalidomide, along with sulfate and glucuronide conjugates. However, the species-specific teratogen thalidomide induces human P450 3A and is activated by P450 3A through the primary human metabolite 5-hydroxythalidomide, leading to its conjugation with nonspecific proteins. The metabolic capacity of polymorphic FMO3 was evaluated based on urine tests for food-derived trimethylamine N-oxide levels. This pharmacokinetic modeling approach, incorporating polymorphic drug-metabolizing enzyme information, could be applied in clinical settings and during computational data-driven evaluations of potential risks associated with a broad spectrum of chemicals.