A ligand-based approach to identify quantitative structure-activity relationships for the androgen receptor

Abstract

We examined the three-dimensional quantitative structure-activity relationship (QSAR) of a group of endogenous and synthetic compounds for the androgen receptor (AR) using comparative molecular field analysis (CoMFA). The goal of these studies was to identify structural features necessary for high binding affinity and optimization of selective androgen receptor modulators (SARMs). A homology model of the AR was used as a scaffold to align six lead compounds that served as templates for alignment of the remaining 116 structures prior to CoMFA modeling. The conventional r(2) and cross-validated q(2) relating observed and predicted relative binding affinity (RBA) were 0.949 and 0.593, respectively. Comparison of predicted and observed RBA for a test set of 10 compounds resulted in an r(2) of 0.954, demonstrating the excellent predictive ability of the model. These integrated homology modeling and CoMFA studies identified critical amino acids for SARM interactions and provided QSAR data as the basis for mechanistic studies of AR structure, function, and design of optimized SARMs.

Figure 1

Alignment points for each set of compounds. (a) Conformation alignment points. Hydroxyflutamide analogues were aligned to the ether-linked SARM, S-4 (compound 5), by the points shown in green. Bicalutamide analogues were aligned using the pink and green points highlighted in S-4. Compounds 24, 29, S-4, and 60 were used as representatives of sulfonyl, thioether, ether, and amide-linked ligands, respectively, for docking. DHT was docked into the homology receptor, and other steroids were then aligned by the points shown in pink to DHT. Compound 85 was docked into the homology receptor, and other tricyclic quinolinones were aligned to this ligand by the points shown in blue. (b) Overlap of ligands generated from field fit. Training and test compounds were aligned to their analogous structure by the points shown in panel a prior to CoMFA analyses.

Figure 2

Plot of residuals and predictions. (a) A plot of the actual vs the predicted pRBA of the training set displays that compounds are closely scattered along the line with a slope of 0.949 and a correlation of 0.974 as calculated by KaleidaGraph 3.5. (b) Plot of the residuals following exclusion of outliers for the training set. (c) A plot of the actual vs the predicted pRBA for the test set demonstrates a correlation of 0.953. (d) Plot of the residuals from the test set.

Figure 3

CoMFA model and ligand interactions. (a) The contour plot shown at 5% contour levels demonstrates the most important areas of ligand interactions. (b) The 30% contour levels allow the contours surrounding the B-ring of compound S-4 to be visualized.

Figure 4

(a) CoMFA model displayed at 5% contour levels emphasizes critical residues for AR binding. (b) CoMFA model displayed at 30% contour levels emphasizes steric hindrance of the B-ring by MET780 and MET787.

Figure 5

Overlap of S-4 (colored by atom), DHT (green), and a tricyclic quinolinone from the docked conformation into the AR homology receptor. Compound S-4 requires an additional space in the receptor compared with steroids and tricyclic quinolinones according to this model.

Figure 6

Overlap of AR crystal structure (colored by atom) and homology model (orange) binding pocket. The major difference between our homology model and the crystal structure is the position of MET780. Docking of bicalutamide analogues to the crystal structure will not occur because of this discrepancy.