Precise prediction of activators for the human constitutive androstane receptor using structure-based three-dimensional quantitative structure-activity relationship methods

The constitutive androstane receptor (CAR, NR1I3) regulates the expression of numerous drug-metabolizing enzymes and transporters. The upregulation of various enzymes, including CYP2B6, by CAR activators is a critical problem leading to clinically severe drug-drug interactions (DDIs). To date, however, few effective computational approaches for identifying CAR activators exist. In this study, we aimed to develop three-dimensional quantitative structure-activity relationship (3D-QSAR) models to predict the CAR activating potency of compounds emerging in the drug-discovery process. Models were constructed using comparative molecular field analysis (CoMFA) based on the molecular alignments of ligands binding to CAR, which were obtained from ensemble ligand-docking using 28 compounds as a training set. The CoMFA model, modified by adding a lipophilic parameter with calculated logD_7.4 (S+logD_7.4), demonstrated statistically good predictive ability (r² = 0.99, q² = 0.74). We also confirmed the excellent predictability of the 3D-QSAR model for CAR activation (r²_pred = 0.71) using seven compounds as a test set for external validation. Collectively, our results indicate that the 3D-QSAR model developed in this study provides precise prediction of CAR activating potency and, thus, should be useful for selecting drug candidates with minimized DDI risk related to enzyme-induction in the early drug-discovery stage.