Systematic Coarse-Grained Models for Molecular Systems Using Entropy

Evangelia Kalligiannaki, Vagelis Harmandaris, Markos A. Katsoulakis2019 · DOI

TL;DR

This paper presents a systematic approach to building coarse-grained (CG) molecular models by leveraging information theory, specifically relative entropy. By minimizing relative entropy, they derive optimal CG potentials that accurately reproduce the structural properties of atomistic simulations, offering a unified framework that connects information-theoretic, structural, and force-matching methods.

What Problem Does This Paper Address?

Simulating complex molecular systems, like materials or biological molecules, is computationally challenging due to the vast range of length and time scales involved. To tackle this, researchers develop "coarse-grained" (CG) models, where groups of atoms are represented as single particles. The core challenge is to derive these CG models systematically and accurately, ensuring they capture the essential structural and dynamical behavior of the original atomistic system. This paper addresses the problem of how to derive these effective CG potentials from detailed atomistic simulations in a principled and systematic way, aiming to bridge the gap between atomistic detail and mesoscopic descriptions.

Methods

The paper focuses on information-theoretic methods, particularly the Relative Entropy (RE), as a principled way to derive optimal CG models. The RE quantifies the difference between the probability distribution of the atomistic system and the CG system. By minimizing the RE, the authors aim to find CG model parameters that make the CG distribution as close as possible to the atomistic one.

They connect this information-theoretic approach to two other established methods:

  • Structural-based methods (e.g., Inverse Boltzmann): These methods rely on matching the radial distribution function (RDF) of the atomistic system.
  • Force Matching (FM): This method aims to minimize the difference between forces calculated from the atomistic system and forces predicted by the CG model.

The paper mathematically demonstrates the equivalence between these methods under certain conditions, particularly for general nonlinear coarse-graining maps. They also introduce Path-Space Force Matching (PSFM) for inferring dynamical models and discuss how Relative Entropy Rate (RER) is used for stationary systems. For quantifying uncertainty, they employ the bootstrap method.

Datasets

The paper uses data generated from molecular dynamics (MD) simulations of the following systems:

  • Bulk Methane: Simulated at 100 K and 80 K.
  • Water: Simulated at 300 K and 1 atm.
  • Polyethylene Melt: Simulated at 450 K.

These simulations provide the "ground truth" atomistic data from which the CG models are derived. No new benchmark datasets are introduced; rather, existing well-characterized molecular systems are used for evaluation.

Key Insights and Ideas

  • Unified Framework: The paper elegantly shows that information-theoretic methods (like Relative Entropy) provide a unifying principle that connects structural-based and force-matching approaches for deriving CG models. This suggests a common underlying theoretical foundation for these seemingly different techniques.
  • Equilibrium and Dynamics: The methods discussed are applicable to both equilibrium properties (structural) and non-equilibrium dynamics, highlighting their versatility.
  • Importance of Information Theory: Relative entropy emerges as a powerful tool for parameterizing CG models, offering a principled way to minimize discrepancies between atomistic and mesoscopic descriptions.
  • Uncertainty Quantification: The application of the bootstrap method to the polyethylene system demonstrates a practical way to estimate confidence intervals for CG potentials, which is crucial for assessing the reliability of the derived models, especially when data is limited.

Key Results

  • Methane: The RE and Inverse Boltzmann (IBI) methods yielded similar effective pair potentials, while Force Matching (FM) showed slight differences.
  • Water: Both RE and FM methods produced pair potentials that could accurately reproduce the target pair correlation function of the atomistic system.
  • Polyethylene: The paper demonstrated the effectiveness of FM for deriving potentials and, importantly, showed that bootstrapping could provide reliable confidence intervals for these potentials, even with a limited number of observations. The PSFM method was shown to be equivalent to FM for equilibrium dynamics.

Limitations and Future Work

  • Asymptotic Equivalence: The paper notes that the equivalence between RE and FM methods is asymptotic. This implies that for finite datasets or specific parameterizations, differences might persist.
  • Computational Cost: While not explicitly detailed as a limitation, deriving these systematic CG models from high-dimensional atomistic simulations is inherently computationally intensive.
  • Future Work: The paper suggests that the presented methodologies can be further applied to more complex systems and that the uncertainty quantification techniques can be extended to provide more robust confidence measures for CG models. The exploration of non-equilibrium dynamics using PSFM is also highlighted as an area for further investigation.

Summary updated

Abstract

The development of systematic coarse-grained mesoscopic models for complex molecular systems is an intense research area. Here we first give an overview of different methods for obtaining optimal parametrized coarse-grained models, starting from detailed atomistic representation for high dimensional molecular systems. We focus on methods based on information theory, such as relative entropy, showing that they provide parameterizations of coarse-grained models at equilibrium by minimizing a fitting functional over a parameter space. We also connect them with structural-based (inverse Boltzmann) and force matching methods. All the methods mentioned in principle are employed to approximate a many-body potential, the (n-body) potential of mean force, describing the equilibrium distribution of coarse-grained sites observed in simulations of atomically detailed models. We also present in a mathematically consistent way the entropy and force matching methods and their equivalence, which we derive for general nonlinear coarse-graining maps. We apply, and compare, the above-described methodologies in several molecular systems: A simple fluid (methane), water and a polymer (polyethylene) bulk system. Finally, for the latter we also provide reliable confidence intervals using a statistical analysis resampling technique, the bootstrap method.
BibTeX
@misc{kalligiannaki2019systematic,
  author = {Evangelia Kalligiannaki and Vagelis Harmandaris and Markos A. Katsoulakis},
  title = {{Systematic Coarse-Grained Models for Molecular Systems Using Entropy}},
  year = {2019},
  doi = {10.3390/ecea-5-06710},
}
Figure