G protein-coupled receptors (GPCRs) are membrane proteins targeted by over one-third of marketed drugs. GPCRs can trigger a wide range of cellular pathways by coupling with diverse effector proteins and undergo large-scale conformational changes across multiple functional states. Many studies have identified the hallmark switches including the outward/inward movement of TM6, the twisting/untwisting of the NPxxY motif, and several other microswitches, to identify conformational substates related to functional transitions. Here, we assembled a molecular dynamics (MD) simulation data set for the A 2A adenosine receptor containing simulations that were initiated under diverse conditions, including different ligands and ligation states, G protein binding scenarios, and membrane environments. By assembling such a diverse set of trajectories, we hoped to maximize the sampling of the receptor’s conformational landscape and to enhance the probability of observing transitions between functionally relevant substates. Dimensionality reduction was performed with the uniform manifold approximation to transform the high-dimensional data into a compact representation, and subsequent clustering (using hierarchical density-based clustering of applications with noise) identified distinct regions of conformation space that correspond to active states, several different intermediate states, and inactive states. Classification and features importance analysis identified—in a fully automated and blind way—specific microswitches that distinguish these states. Our analysis reproduced previously established features and microswitches, while also revealing a possibly novel microswitch—a specific region on TM2. These findings suggest that well-chosen dimensionality reduction and clustering methods are powerful tools for understanding functional transitions in proteins.
Ji et al. (Sun,) studied this question.