The bottleneck in enhanced sampling lies in finding collective variables that effectively accelerate protein conformational changes; true reaction coordinates that accurately predict the committor are the well-recognized optimal choice. However, identifying them requires unbiased natural reactive trajectories, which, paradoxically, require effective enhanced sampling. Using the generalized work functional method, we uncover that true reaction coordinates control both conformational changes and energy relaxation, enabling us to compute them from energy relaxation simulations. Biasing true reaction coordinates accelerates conformational changes and ligand dissociation in PDZ2 domain and HIV-1 protease by 105 to 1015-fold. The resulting trajectories follow natural transition pathways, enabling efficient generation of unbiased reactive trajectories. In contrast, biased trajectories from empirical collective variables display non-physical features. Furthermore, our method uses a single protein structure as input, enabling predictive sampling of conformational changes. These findings unlock access to a broader range of protein functional processes in molecular dynamics simulations.
Read full abstract