Locating ligand binding sites in G-protein coupled receptors using combined information from docking and sequence conservation

Abstract

G-protein coupled receptors (GPCRs) are the largest protein family of drug targets.Detailed mechanisms of binding are unknown for many important GPCR-ligand pairs due to the difficulties of GPCR recombinant expression, biochemistry, and crystallography. We describe our new method, ConDock, for predicting ligand binding sites in GPCRs using combined information from surface conservation and docking starting from crystal structures or homology models. We demonstrate the effectiveness of ConDock on well-characterized GPCRs such as the 2 adrenergic and A2A adenosine receptors. We also demonstrate that ConDock successfully predicts ligand binding sites from high-quality homology models. Finally, we apply ConDock to predict ligand binding sites on a structurally uncharacterized GPCR, GPER. GPER is the Gprotein coupled estrogen receptor, with four known ligands: estradiol, G1, G15, and tamoxifen.ConDock predicts that all four ligands bind to the same location on GPER, centered on L119, H307, and N310; this site is deeper in the receptor cleft than predicted by previous studies. We compare the sites predicted by ConDock and traditional methods that utilize information from surface geometry, surface conservation, and ligand chemical interactions. Incorporating sequence conservation information in ConDock overcomes errors introduced from physics-based scoring functions and homology modeling.

Computational Approaches for the Discovery of GPER Targeting Compounds

“…A major strength of the algorithm used in GPCR-I-TASSER is that it was specifically devised for the prediction of GPCRs and, as such, it works with a dedicated knowledge-based force field as a guide for an accurate assembly of the receptor structure. A number of works ( 25 , 38 , 42 , 43 ) have adopted this computational tool to design high-quality GPER models (see again Figure 1 ). High-quality and tailored software, together with the growing number of experimentally-determined GPCR structures that can be used as GPER templates, are contributing to make homology modeling algorithms more accurate in predicting the receptor structure.…”