All-atom molecular dynamics simulations are a powerful tool for gaining insights into chemical systems. Due to the computational cost of all-atom models, low-resolution coarse-grained (CG) models are widely adopted for studying larger, longer length scale systems such as polymers and biomolecules. However, there exist many unanswered questions regarding the first step in constructing a CG model - i.e., how best to represent a system in reduced detail. In this study, we develop a statistical physics framework for investigating these questions. We adopt the Gaussian Network Model (GNM) as a simple model for proteins. Given the GNM, we can analytically evaluate all properties of a CG model constructed using any reduced resolution representation. We assess the quality of each CG representation using a variety of metrics which describe the preservation of different aspects of the underlying high-resolution model. We implement Monte Carlo (MC) sampling to sample the space of possible CG representations of the system with a fixed number of CG sites. We develop a new sampling algorithm that allows for the number of atoms in each CG site to fluctuate, while the number of sites remains fixed. We find that this new algorithm determines higher quality representations than a simpler algorithm that required all sites to be the same size. We also find that the ability of the model to preserve the slow-low frequency modes is anti-correlated with the amount of information preserved by the model at low resolutions but un-correlated at higher resolutions.

Investigation of the space of coarse-grained model representations

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.