An Open Source in silico Prediction Model of Hepatic Intrinsic Clearance for in vitro to in vivo Extrapolation

March 20, 2020

Download Poster PDF: Clearance Tool SOT 2020 Poster PDF

Authors: Jeremy Fitzpatrick, Pankajini Mallick, Kevin Bronson, Patrick McMullen

Chemical risk assessment is moving away from animal testing replacing it with computational and in vitro approaches in risk assessment. Dosimetry, by determining the amounts and key rates for distribution of a chemical in the body, is essential for the use of new approach methodologies in risk-based decision making. Use of models such as high-throughput in vitro to in vivo extrapolation (HT-IVIVE) has allowed dosimetry considerations to be more readily incorporated earlier in safety decision making processes. To keep pace with de novo computational approaches for estimating hazard, new tools are needed that can provide reasonable estimates of chemical clearance without requiring wet-lab experimentation and analytical chemistry. To this end we have developed a tool—using published metabolism information and physicochemical properties—to estimate a key determinant of compound dosimetry: intrinsic clearance. This tool allows a user to input a novel structure by hand or a series of structures using a chemical structure file such as a SMILES or SDF file. A prediction is then generated using a nearest neighbors model. The prediction is reported to be in or out of the model’s domain based on its structural similarity to its nearest neighbors. The domain was initially determined by using a leave-one-out validation of a training set of 426 chemicals. The optimized model was then used to generate predictions for a test of 107 chemicals.