In Silico Prediction of Toxicokinetic Parameters
In Silico Prediction of Toxicokinetic Parameters for Environmentally Relevant Chemicals for Risk-Based Prioritization
Toxicokinetic (TK) models can address an important component of chemical risk assessments by helping bridge the gap between chemical exposure and measured toxicity endpoints. The metabolic clearance rate (CLint) and fraction of a chemical unbound by plasma proteins (Fub) are critical TK parameters, accounting for aspects of the distribution, metabolism and excretion that determine in vivo tissue concentrations. Yet, limited data are available for these two parameters for environmentally relevant chemicals, including approximately 8000 chemicals with in vitro bioactivity data collected by Tox21. Quantitative structure-activity relationships (QSAR) for CLint and Fub were developed with in vitro assay data for both pharmaceuticals and chemicals in the ToxCast screening initiative using machine learning algorithms and open source descriptors. The models were shown to offer reliable in silico predictions of CLint and Fub for a diverse array of chemicals within the applicability domains. Incorporating the QSARs into TK models allowed a high throughput risk-based prioritization scheme informed by the margin between bioactive doses and human exposure. These QSAR models aid in the high-throughput identification and prioritization of those chemicals with the highest probability of triggering adverse outcomes.
The presented work is that of the authors and does not necessarily represent U.S. EPA views or policies.