Because they eliminate unnecessary degrees of freedom, coarse-grained (CG) models enable studies of phenomena that are intractable with more detailed models. For the same reason, the effective potentials that govern CG degrees of freedom incorporate entropic contributions from the eliminated degrees of freedom. Consequently, these effective potentials demonstrate limited transferability and provide a poor estimate of atomic energetics. Here, we propose a simple dual-potential approach that combines “structure-based” and “energy-based” variational principles to determine effective potentials that model free energies and potential energies, respectively, as a function of the CG configuration. We demonstrate this approach for 1-site CG models of water and methanol. We accurately sample configuration space by performing simulations with the structure-based potential. We accurately estimate average atomic energies by postprocessing the sampled configurations with the energy-based potential. Finally, the difference between the two potentials predicts a qualitatively accurate estimate for the temperature dependence of the structure-based potential.

The dual-potential approach promises coarse-grained (CG) models that accurately reproduce both structural and energetic properties, while simultaneously providing predictive estimates for the temperature-dependence of the effective CG potentials. In this work, we examine the dual-potential approach for implicit solvent CG models that reflect large entropic effects from the eliminated solvent. Specifically, we construct implicit solvent models at various resolutions, R, by retaining a fraction 0.10 ≤ R ≤ 0.95 of the molecules from a simple fluid of Lennard-Jones spheres. We consider the dual-potential approach in both the constant volume and constant pressure ensembles across a relatively wide range of temperatures. We approximate the many-body potential of mean force for the remaining solutes with pair and volume potentials, which we determine via multiscale coarse-graining and self-consistent pressure-matching, respectively. Interestingly, with increasing temperature, the pair potentials appear increasingly attractive, while the volume potentials become increasingly repulsive. The dual-potential approach not only reproduces the atomic energetics but also quite accurately predicts this temperature-dependence. We also derive an exact relationship between the thermodynamic specific heat of an atomic model and the energetic fluctuations that are observable at the CG resolution. With this generalized fluctuation relationship, the approximate CG models quite accurately reproduce the thermodynamic specific heat of the underlying atomic model.

Due to their computational efficiency, coarse-grained (CG) models are widely adopted for modeling soft materials. As a consequence of averaging over atomistic details, the effective potentials that govern the CG degrees of freedom vary with temperature and density. This state-point dependence not only limits their range of validity but also presents difficulties when modeling thermodynamic properties. In this work, we systematically examine the temperature- and density-dependence of effective potentials for 1-site CG models of liquid ethane and liquid methanol. We employ force-matching and self-consistent pressure-matching to determine pair potentials and volume potentials, respectively, that accurately approximate the many-body potential of mean force (PMF) at a range of temperatures and densities. The resulting CG models quite accurately reproduce the pair structure, pressure, and compressibility of the corresponding all-atom models at each state point for which they have been parameterized. The calculated pair potentials vary quite linearly with temperature and density over the range of liquid state points near atmospheric pressure. These pair potentials become increasingly repulsive both with increasing temperature at constant density and also with increasing density at constant temperature. Interestingly, the density-dependence appears to dominate, as the pair potentials become increasingly attractive with increasing temperature at constant pressure. The calculated volume potentials determine an average pressure correction that also varies linearly with temperature, although the associated compressibility correction does not. The observed linearity allows for predictions of pair and volume potentials that quite accurately model these liquids in both the constant NVT and constant NPT ensembles across a fairly wide range of temperatures and densities. More generally, for a given CG configuration and density, the PMF will vary linearly with temperature over the temperature range for which the entropy associated with the conditioned distribution of atomic configurations remains constant.

We present the BOCS toolkit as a suite of open source software tools for parametrizing bottom-up coarse-grained (CG) models to accurately reproduce structural and thermodynamic properties of high-resolution models. The BOCS toolkit complements available software packages by providing robust implementations of both the multiscale coarse-graining (MS-CG) force-matching method and also the generalized-Yvon-Born-Green (g-YBG) method. The g-YBG method allows one to analyze and to calculate MS-CG potentials in terms of structural correlations. Additionally, the BOCS toolkit implements an extended ensemble framework for optimizing the transferability of bottom-up potentials, as well as a self-consistent pressure-matching method for accurately modeling the pressure equation of state for homogeneous systems. We illustrate these capabilities by parametrizing transferable potentials for CG models that accurately model the structure, pressure, and compressibility of liquid alkane systems and by quantifying the role of many-body correlations in determining the calculated pair potential for a one-site CG model of liquid methanol.

The high concentration of nucleic acids and positively charged proteins in the cell nucleus provides many possibilities for complex coacervation. We consider a prototypical mixture of nucleic acids together with the polycationic C-terminus of histone H1 (CH1). Using a minimal coarse-grained model that captures the shape, flexibility, and charge distributions of the molecules, we find that coacervates are readily formed at physiological ionic strengths, in agreement with experiment, with a progressive increase in local ordering at low ionic strength. Variation of the positions of charged residues in the protein tunes phase separation: for well-mixed or only moderately blocky distributions of charge, there is a modest increase of local ordering with increasing blockiness that is also associated with an increased propensity to phase separate. This ordering is also associated with a slowdown of rotational and translational diffusion in the dense phase. However, for more extreme blockiness (and consequently higher local charge density), we see a qualitative change in the condensed phase to become a segregated structure with a dramatically increased ordering of the DNA. Naturally occurring proteins with these sequence properties, such as protamines in sperm cells, are found to be associated with very dense packing of nucleic acids.