Using Committor and its Distribution to Assess the Convergence of Free Energy Calculations

Abstract

Molecular simulation has been extensively used to calculate free energy associated with any molecular phenomenon in order to understand the thermodynamics and kinetics of the process. When the energetic barrier between the initial and the final stage is large, biasing collective variables (cv) that are some combination of atomic coordinates, allows sampling of all representative configurations and accurate calculation of free energy profiles. In this work, using all atom molecular dynamics simulation with biasing schemes like well-tempered metadynamics, umbrella sampling and replica exchange umbrella sampling, we calculate the free energy profile for translocation of small molecules through a model 1,2-dioleoyl-sn-glycero-3-phosphocholinebilayer as a function of distance from the bilayer center. We find that only for neutral solutes, the free energy calculation is straightforward. Using the committor analysis, we find that non-converged profiles result due to insufficient sampling of the transition state. We also demonstrate the usage of acceptance probabilities in a replica exchange calculation as a way to identify sampling problems. Based on our results, we provide useful guidance to detect and overcome convergence failure to infer useful physical behavior of a system.