Optimization of coarse-grained models: matching probability density in conformational space

TL;DR

This paper introduces a novel methodology for constructing coarse-grained models by minimizing the difference in conformational probability densities between CG and all-atomic models, ensuring accurate physical property reproduction.

Contribution

It proposes a new variational framework using basis functions to optimize CG force fields by matching probability densities and extending to pressure reproduction, verified on a water model.

Findings

Effective CG models can accurately reproduce equilibrium properties.

The methodology successfully constructs a one-site CG water model.

The approach generalizes to include pressure matching.

Abstract

Coarse-Graining (CG) models are low resolution approximation of high resolution models, such as all-atomic (AA) models. An effective CG model is expected to reproduce equilibrium values of sufficient physical quantities of its AA model, which requires to match the equilibrium probability density of the CG model to that of the AA model in conformational space. The present work proposes for constructing effective CG models a novel methodology that aims at minimizing the distance between CG model and AA model. The distance is defined as a functional of conformational probability densities in CG and AA models and further expanded by ensemble averages of a set of sufficient and independent basis functions. An orthogonalization strategy is adopted to get the independent basis functions from sufficiently preselected interesting physical quantities of the system. Two variational methods are developed to optimize parameters of effective CG force field by minimizing the functional of probability densities, are then generalized so that the CG model also reproduce the pressure of AA model. The general CG framework is verified in constructing one-site CG water from TIP3P water model.