CoWERA: A temporal coherence guided binless resampling algorithm for weighted-ensemble based estimation of rare-event kinetics

Rare-event processes in molecular systems, such as protein folding, conformational switching, and ligand (un)binding, often lie far beyond the timescales accessible to straightforward molecular dynamics simulations. Weighted ensemble (WE) simulations can accelerate sampling without biasing the underlying dynamics, but their practical efficiency is strongly governed by how trajectories are resampled. Here, we introduce CoWERA (Coherence-based Weighted Ensemble Resampling Algorithm), a binless, targeted WE resampling strategy that prioritizes trajectories using "temporal coherence," i.e., the persistence of forward progress toward a target, rather than instantaneous position in collective-variable space. CoWERA defines a "trajectory intensity" from signed progress over an adaptively chosen history window and uses this metric to guide cloning of productive walkers and merging of unproductive ones while conserving statistical weights. We benchmark CoWERA on chignolin and Trp-cage miniproteins and show that coherence-guided resampling yields reactive events rapidly along with faster stabilization of rate estimates and reduced run-to-run variability as compared to conventional binned WE and a proximity-based targeted WE baseline. For Trp-cage, CoWERA reproduces reference folding and unfolding mean first-passage times while requiring substantially less aggregate simulation time, demonstrating that incorporating history-dependent temporal coherence into resampling decisions can improve the robustness and efficiency of WE-based simulations for estimating rare event kinetics.