Ultra-large virtual screening unveils potent agonists of the neuromodulatory orphan receptor GPR139.

Abstract

The orphan G protein-coupled receptor (GPCR) GPR139 attracts interest as a target for neuropsychiatric disorders. Whereas the physiological functions of GPR139 remain elusive, a high-resolution receptor structure is now available. To assess whether structural information enables ligand discovery, we computationally dock 235 million compounds to the GPR139 binding site. Of 68 top-ranked compounds evaluated experimentally, five are full agonists with potencies ranging from 160 nM to 3.6 µM. Structure-guided optimization identifies one of the most potent GPR139 agonists, and a cryo-EM structure of the receptor-ligand complex confirms the predicted binding mode. Functional characterization provides insights into GPR139 signalling, and one agonist elicits behavioural effects in mice. We also explore the potential to replace experimental structure determination with the deep-learning method AlphaFold3, revealing a limited capability of artificial intelligence to model receptor-ligand interactions for understudied GPCRs. The results demonstrate how high-resolution GPCR structures combined with large-library docking can accelerate drug discovery.