Orphan GPCRs

Over 120 GPCRs (60 in Class A) remain classified as orphans [123] , that is, their endogenous ligands (and in many cases, synthetic ligands) have not been identified. There is a great interest in the functional and structural analysis of orphan receptors, which often leads to identification of new therapeutic targets. Limited structural information and lack of ligand binding data make orphan GPCRs especially hard targets for structure-based modeling.

Chemogenomics approaches allow clustering of orphan GPCRs with well-described receptors and attempt to extrapolate known SAR correlations between ligand activity and receptor/ligand structure to the orphan receptor [ 163] . Such approaches are not always useful in the search for endogenous ligands, as the pharmacological and phylogenetic classification of GPCRs is often paradoxically different. For example, GPCRs that recognize small molecule ligands versus peptides are not clearly separated in a phylogenetic tree. However, a detailed analysis of residue conservation in the 7TM binding cavity can greatly facilitate a focused search for synthetic, often allosteric, agonists and antagonists [3, 4] . Knowledge-based models have been trained on large sets of GPCR ligand binding data using machine learning techniques (e.g., support vector machines, or SVM) and generalized protein sequence and "ligand feature" representation [164]. Note that these bioinformatics approaches could greatly benefit from understanding the 3D context of the binding pocket, even for relatively low-resolution orphan GPCR models. For example, the accuracy of the knowledge-based models may be improved by using information about side chains lining the binding pocket and spatial distances between them. Though the direct application of small molecule docking to GPCR deorphanization has yet to be reported, accumulation of relevant structural templates and improved modeling techniques can eventually lead to more accurate orphan GPCR models for use in in silico ligand prediction. Induced fit modeling techniques, such as SCARE, may prove especially useful for predicting the bound conformations of new ligands [50].

0 0

Post a comment