Novel CAR inverse agonist candidates: a combination of docking–QSAR inhibition activity modeling of a series of CINPA1 analogs as CAR inverse agonist by using BGSA-BRANN method

Abstract

In biological research, increasing the effectiveness of anticancer drugs and attenuating multidrug resistance are of prime importance. In this study, a combination of docking–quantitative structure activity relationship (QSAR) method is implemented for modeling and predicting inhibition activity (pIC50) of CINPA1 analogs as the constitutive androstane receptor (CAR) inverse agonists. Docking studies were used to find the optimal ligands conformation and interactions as well as their position and orientation within the binding site. Then, the docking-derived conformations and orientations were implemented to calculate QSAR molecular descriptors. A data set containing 50 molecules with known pIC50 was divided into six series of training and test sets, each including 40 and 10 molecules, respectively. Binary gravitational search algorithm and Bayesian regularization-based neural networks (BRANN) were applied to develop a general model using the most informative descriptors. The results of external and internal cross-validation tests in conjunction with Y-randomization confirm predictive ability, robustness and effectiveness of the generated models. It is revealed that hydrophobic interactions, number of nitrogen atoms and cation-π interactions play important roles in the CAR inhibition activity of the agents. Finally, by applying the best model, several novel CAR inverse agonist candidates were proposed for further experimental studies.