Free Energy Landscape Perturbation for Estimating Mutation-Induced Population Shifts from Reference Molecular Dynamics Ensembles

Abstract

Mutations can alter protein function by redistributing probability among conformational states, even when structural changes are modest. Directly estimating a perturbed conformational free energy landscape with molecular dynamics is expensive because each mutation requires a new converged simulation. Here, we introduce Free Energy Landscape Perturbation (FELP), a state-local reweighting framework that estimates mutation-induced population shifts from a reference ensemble. FELP partitions a reference trajectory into discrete microstates, evaluates restrained reference-to-perturbed energy differences for configurations sampled within each state, and combines spatially regularized state-local Zwanzig factors with the reference state probabilities to estimate the perturbed state distribution. FELP estimates how a perturbation changes the relative probabilities of conformational basins and reports effective-sample and overlap diagnostics that identify when the predicted shift lacks support in the sampled reference ensemble. We describe the FELP estimator, its implementation, and its performance on capped dipeptides, chignolin, T4 lysozyme, and the EGFR kinase domain. FELP is intended for perturbations that reshape populations within a sufficiently sampled reference landscape, providing a practical bridge between local alchemical energy evaluation and ensemble-level mechanistic interpretation.