pdCSM-GPCR: In silico prediction of GPCR ligands



Abstract: The G protein coupled receptors superfamily is one of the most widely class of proteins screened for ligands. Despite the great effort directed towards the gpcr ligand discovery, many endogenous ligands still remain unknown (orphan receptors) and there are still leakage of safe and effective drug for many GPCR of medical interest. With recent advances in computational power, and machine learning algorithms, prediction of ligand affinity is getting more and more feasible. We take advantage of it to discovery new ligands for GPCRs through assessment of ligand bioactivities. This can guide rational experimentation in finding and validating novel ligands for GPCRs.

Our approach is called pdCSM-GPCR, and relies on graph-based signatures. These encode distance patterns between atoms and are used to represent the small molecule and to train predictive models. Here we present a web server which provides a reliable and cost-free platform to rapidly screen ligands for GPCR.