Agonist Binding to Chemosensory Receptors: A Systematic Bioinformatics Analysis

Abstract

Human G-protein coupled receptors (hGPCRs) constitute a large and highly pharmaceutically relevant membrane receptor superfamily. About half of the hGPCRs' family members are chemosensory receptors, involved in bitter taste and olfaction, along with a variety of other physiological processes. Hence these receptors constitute promising targets for pharmaceutical intervention. Molecular modeling has been so far the most important tool to get insights on agonist binding and receptor activation. Here we investigate both aspects by bioinformatics-based predictions across all bitter taste and odorant receptors for which site-directed mutagenesis data are available. First, we observe that state-of-the-art homology modeling combined with previously used docking procedures turned out to reproduce only a limited fraction of ligand/receptor interactions inferred by experiments. This is most probably caused by the low sequence identity with available structural templates, which limits the accuracy of the protein model and in particular of the side-chains' orientations. Methods which transcend the limited sampling of the conformational space of docking may improve the predictions. As an example corroborating this, we review here multi-scale simulations from our lab and show that, for the three complexes studied so far, they significantly enhance the predictive power of the computational approach. Second, our bioinformatics analysis provides support to previous claims that several residues, including those at positions 1.50, 2.50, and 7.52, are involved in receptor activation.

Keywords: G-protein coupled receptor; bioinformatics; bitter taste receptor; chemosensory receptor; homology modeling; molecular docking; molecular mechanics/coarse grained simulations; odorant receptor.

Figure 1

Precision and Recall plots for the predictions of hChem-GPCR/agonist complexes. (A-C) show the HADDOCK (Dominguez et al., 2003), AutoDock Vina (Trott and Olson, 2010), and Glide (Friesner et al., 2004) docking predictions, respectively. The abbreviations used for the hChem-GPCR/agonist complexes are listed in Table 2. Bioinformatics/Docking-based predictions are shown as circles, colored according to the two performance metrics: dark blue (0 precision, 0 recall), red (precision 1, low recall), yellow (low precision, recall 1) and cyan (all the rest, with intermediate precision and recall values). In panel A, MM/CG simulation results (Marchiori et al., ; Sandal et al., 2015), started from Haddock docking complexes, are displayed as colored triangles.

Figure 2

Scheme showing the definition of true positive (TP), false positive (FP), true negative (TN) and false negative (FN) residues used in this study. Comparison of predicted residues with experimental data (EC₅₀ values) is performed on the basis of both a distance cut-off (5.5 Å) and a chemical definition (i.e., presence or absence of a canonical protein/ligand interaction).