ManifoldEM is a computational approach that employs nonlinear dimensionality reduction to recover continuous conformational landscapes and reconstructs 3D molecular movies and corresponding free-energy surfaces from cryo-EM images. Its strength lies in providing a continuous spectrum of states and associated free-energy surface but fails to provide time-resolved kinetics. In contrast, milestoning simulations compute both kinetics and free-energy profiles for complex molecular processes by partitioning the phase space into intermediate states separated by milestone boundaries. Short trajectories are then run within milestones, and transition statistics are collected to obtain rate constants and thermodynamic state populations. This approach provides time-resolved information but requires a priori definition of appropriate milestones. We thereby introduce a hybrid workflow that integrates ManifoldEM with milestoning simulations to combine the respective strengths of each method. First, ManifoldEM is applied to cryo-EM data sets to construct a low-dimensional conformational free-energy surface. Optimal milestones are then placed along the manifold by minimizing the total cycle time, defined as the mean round-trip transition time between adjacent milestones, estimated from the underlying free-energy and diffusion profiles using the Smoluchowski equation. The resulting milestones, denser near steep barriers and sparser in flat regions, define discrete states spanning the dominant conformational transition coordinate. These milestones serve as input for atomic simulations to obtain a refined free-energy profile and transition kinetics. Because milestones were chosen based on the ManifoldEM landscape, simulations focus on the most relevant conformational transitions, avoiding wasted effort on nonessential degrees of freedom. This integrated ManifoldEM-milestoning workflow yields conformational free-energy landscapes and kinetics of dominant motions, bridging static ensembles with time-resolved dynamics. Rapid ManifoldEM analysis, followed by refined milestone simulations, provides accurate rates and free-energy profiles for biomolecular transitions.
Anupam Anand Ojha (Sun,) studied this question.