Approximating nonequilibrium processes using a collection of surrogate diffusion models

TL;DR

This paper introduces a method to approximate nonequilibrium peptide dynamics using surrogate diffusion models, enabling prediction of work distributions from limited simulation data.

Contribution

The study applies surrogate process approximation (SPA) to model peptide unfolding and refolding dynamics, revealing structural information beyond the monitored reaction coordinate.

Findings

SPA models contain orthogonal structural information

Work distribution shapes can be predicted with limited data

Functional data analysis effectively correlates SPA models with work done

Abstract

The surrogate process approximation (SPA) is applied to model the nonequilibrium dynamics of a reaction coordinate (RC) associated with the unfolding and refolding processes of a deca-alanine peptide at 300 K. The RC dynamics, which correspond to the evolution of the end-to-end distance of the polypeptide, are produced by steered molecular dynamics (SMD) simulations and approximated using overdamped diffusion models. We show that the collection of (estimated) SPA models contain structural information "orthogonal" to the RC monitored in this study. Functional data analysis ideas are used to correlate functions associated with the fitted SPA models with the work done on the system in SMD simulations. It is demonstrated that the shape of the nonequilibrium work distributions for the unfolding and refolding processes of deca-alanine can be predicted with functional data analysis ideas using a relatively small number of simulated SMD paths for calibrating the SPA diffusion models.