Collaborative multi-agent intelligence uncovers subtype-selective allosteric sites at GPCR-lipid interaction interface

Abstract

Closely related G protein-coupled receptor (GPCR) subtypes often share highly conserved orthosteric pockets, making subtype-selective ligand development challenging. Here, we developed a five-agent workflow to systematically identify divergent protein-membrane-interface sites across class A GPCRs and exploit them for selective allosteric ligand discovery. By combining dMaSIF-derived surface fingerprints with Ballesteros-Weinstein (BW) position alignment, we compared structurally equivalent membrane-facing regions and identified the three most divergent hotspots for each of 163 receptor pairs. These regions showed substantial spatial overlap with experimentally characterized allosteric sites. Paired target-off-target screening of one million lead-like compounds, followed by detail-mode redocking and multi-seed consistency filtering, yielded 352 receptor-pair-specific candidates corresponding to 344 unique compounds across 104 receptor pairs. These candidates, together with their divergent sites and predicted selectivity profiles, were integrated into a searchable database. Our findings establish a scalable strategy for translating GPCR membrane-interface divergence into precise allosteric sites and testable subtype-selective ligand candidates.