Prediction of cytotoxicity of polycyclic aromatic hydrocarbons from first principles

Ligand-specific binding interactions of xenobiotics with receptor proteins form the basis of cytotoxicity-based hazard assessment. Computational approaches enable predictive hazard assessment for a large number of chemicals in a high-throughput manner, minimizing the use of animal testing. However, in silico models for predicting mechanisms of toxic actions and potencies are difficult to develop because toxicity datasets or comprehensive understanding of the complicated kinetic process of ligand-receptor interactions are needed for model development. In this study, a directional reactive binding factor (DRBF) model based on first principles was used to predict cytotoxicity potencies of agonists of the aryl hydrocarbon receptor (AhR) for 16 different polycyclic aromatic hydrocarbons (PAHs). Molecular dynamics were simulated by accounting for the directional configuration factor toward receptor protein and the factor of binding to the Per-Arnt-Sim (PAS) domain. When comparing the experimental results of toxic potencies from in vitro bioassays with the predictions among two different in silico models, including quantitative structure-activity relationship (QSAR) and molecular docking models, the DRBF model exhibited the highest model performance (R2 = 0.90 and p