Quantitative structure–activity relationships for small non-peptide antagonists of CXCR2: Indirect 3D approach using the frontal polygon method

Abstract

The chemokine receptor, CXCR2, plays an important role in recruiting granulocytes to sites of inflammation and has been proposed as an important therapeutic target. A number of CXCR2 antagonists have been synthesized and evaluated; however, quantitative structure–activity relationship (QSAR) models have not been developed for these molecules. Most CXCR2 antagonists can be grouped into four related categories: N, N′-diphenylureas, nicotinamide N-oxides, quinoxalines, and triazolethiols. Based on these categories, we developed a QSAR model for 59 nonpeptide antagonists of CXCR2 using a partial 3D comparison of the antagonists with local fingerprints obtained from rigid and flexible fragments of the molecules. Each compound was represented by calculated structural descriptors that encoded atomic charge, molar refraction, hydrophobicity, and geometric features. We obtained good conventional R 2 coefficients, high leave-one-out cross-validated values for the whole dataset ( R cv 2 = 0.785 ) , as well as for the dataset divided into subsets of triazolethiol derivatives ( R cv 2 = 0.821 ) and joint subset of N′-diphenylureas, nicotinamide N-oxides, N, N ′-diphenylureas, and quinoxaline derivatives and quinoxalines derivatives ( R cv 2 = 0.766 ) , indicating a good predictive ability and robustness of the model. Additionally, charge distribution was found to be a significant contributor in modeling whole dataset. Using our model, structural fragments (submolecules) responsible for the antagonist activity were also identified. These data suggest the QSAR models developed here may be useful in guiding the design of CXCR2 antagonists from molecular fragments.