Activation mechanism of class A GPCRs: machine learning analysis of experimental structural databases

Recent advances in cryo-electron microscopy and cryo-electron tomography have dramatically increased the number of class A G protein-coupled receptor (GPCR) structures, especially in previously inaccessible G protein-bound, active-like conformations. The increased structural diversity provides a unique opportunity to explore the conformational landscape underlying GPCR activation. To this end, we developed a machine learning (ML) framework that utilizes experimental structural data to elucidate the activation dynamics of class A GPCRs. We find that receptors can populate both inactive and active-like conformations even in the absence of ligand or G protein, providing a structural basis for agonist-free basal activity. Agonist binding shifts this conformational ensemble towards the active state but does not fully stabilize it. Instead, a stable active state is only established upon G protein binding, which locks the receptor in its active conformation. These results support a hybrid activation mechanism in which ligand binding follows conformational selection, while transducer engagement is governed by induced fit. Beyond clarifying class A GPCR activation, the openly available and modifiable ML framework provides a practical tool for analyzing newly determined structures, investigating the mechanisms of action of other GPCR classes and protein families, and guiding structure-based drug discovery in important pharmacological superfamilies.