Construction and Refinement of Coarse-Grained Models

TL;DR

This paper presents a comprehensive scheme combining coarse-grained modeling, weighted ensemble dynamics simulations, and cluster analysis to efficiently study the dynamical and thermodynamical properties of complex systems without prior state knowledge.

Contribution

It introduces a novel integrated approach for constructing and refining coarse-grained models and analyzing system dynamics through ensemble simulations and clustering.

Findings

Efficient optimization of coarse-grained models based on ensemble mean distances.

Generation of equilibrium trajectories from CG models using weighted ensemble dynamics.

Identification of stable states and transition pathways via cluster analysis.

Abstract

A general scheme, which includes constructions of coarse-grained (CG) models, weighted ensemble dynamics (WED) simulations and cluster analyses (CA) of stable states, is presented to detect dynamical and thermodynamical properties in complex systems. In the scheme, CG models are efficiently and accurately optimized based on a directed distance from original to CG systems, which is estimated from ensemble means of lots of independent observable in two systems. Furthermore, WED independently generates multiple short molecular dynamics trajectories in original systems. The initial conformations of the trajectories are constructed from equilibrium conformations in CG models, and the weights of the trajectories can be estimated from the trajectories themselves in generating complete equilibrium samples in the original systems. CA calculates the directed distances among the trajectories and groups their initial conformations into some clusters, which correspond to stable states in the original systems, so that transition dynamics can be detected without requiring a priori knowledge of the states.