Enhanced sampling of protein conformational states for dynamic cross‐docking within the protein‐protein docking server SwarmDock

The formation of specific protein‐protein interactions is often a key to a protein's function. During complex formation, each protein component will undergo a change in the conformational state, for some these changes are relatively small and reside primarily at the sidechain level; however, others may display notable backbone adjustments. One of the classic problems in the protein‐docking field is to be able to a priori predict the extent of such conformational changes. In this work, we investigated three protocols to find the most suitable input structure conformations for cross‐docking, including a robust sampling approach in normal mode space. Counterintuitively, knowledge of the theoretically best combination of normal modes for unbound‐bound transitions does not always lead to the best results. We used a novel spatial partitioning library, Aether Engine (see Supplementary Materials), to efficiently search the conformational states of 56 receptor/ligand pairs, including a recent CAPRI target, in a systematic manner and selected diverse conformations as input to our automated docking server, SwarmDock, a server that allows moderate conformational adjustments during the docking process. In essence, here we present a dynamic cross‐docking protocol, which when benchmarked against the simpler approach of just docking the unbound components shows a 10% uplift in the quality of the top docking pose.