Prediction of ligand-receptor pharmacological activities using a combined docking and machine learning approach
G protein coupled receptors (GPCRs) are valuable therapeutic targets for many diseases. A central question of GPCR drug discovery is to understand what determines the agonism or antagonism of ligands which bind them. Ligands exert their action via the interactions in the ligand binding pocket. We hypothesised that there is a common set of receptor interactions made by ligands of diverse structures that mediate their action and that among a large dataset of different ligands, the functionally important interactions will be over-represented. We computationally docked ~2700 known {beta}2AR ligands to multiple {beta}2AR structures, generating ca 75,000 docking poses and predicted all atomic interactions between the receptor and the ligand. We used machine learning (ML) techniques to identify specific interactions that correlate with the agonist or antagonist activity of these ligands. The interpretation of ML analysis in human understandable form allowed us to construct an exquisitely detailed structure-activity relationship that identifies small changes to the ligands that invert their activity and thus helps to guide the drug discovery process. This approach can be readily applied to any drug target.