The multiscale coarse-graining method. I. A rigorous bridge between atomistic and coarse-grained models

Coarse-grained (CG) models provide a computationally efficient method for rapidly investigating the long time- and length-scale processes that play a critical role in many important biological and soft matter processes. Recently, Izvekov and Voth introduced a new multiscale coarse-graining (MS-CG) method [J. Phys. Chem. B 109, 2469 (2005); J. Chem. Phys. 123, 134105 (2005)] for determining the effective interactions between CG sites using information from simulations of atomically detailed models. The present work develops a formal statistical mechanical framework for the MS-CG method and demonstrates that the variational principle underlying the method may, in principle, be employed to determine the many-body potential of mean force (PMF) that governs the equilibrium distribution of positions of the CG sites for the MS-CG models. A CG model that employs such a PMF as a "potential energy function" will generate an equilibrium probability distribution of CG sites that is consistent with the atomically detailed model from which the PMF is derived. Consequently, the MS-CG method provides a formal multiscale bridge rigorously connecting the equilibrium ensembles generated with atomistic and CG models. The variational principle also suggests a class of practical algorithms for calculating approximations to this many-body PMF that are optimal. These algorithms use computer simulation data from the atomically detailed model. Finally, important generalizations of the MS-CG method are introduced for treating systems with rigid intramolecular constraints and for developing CG models whose equilibrium momentum distribution is consistent with that of an atomically detailed model.     

The multiscale coarse-graining method. II. Numerical implementation for coarse-grained molecular models

The multiscale coarse-graining (MS-CG) method [S. Izvekov and G. A. Voth, J. Phys. Chem. B 109, 2469 (2005); J. Chem. Phys. 123, 134105 (2005)] employs a variational principle to determine an interaction potential for a CG model from simulations of an atomically detailed model of the same system. The companion paper proved that, if no restrictions regarding the form of the CG interaction potential are introduced and if the equilibrium distribution of the atomistic model has been adequately sampled, then the MS-CG variational principle determines the exact many-body potential of mean force (PMF) governing the equilibrium distribution of CG sites generated by the atomistic model. In practice, though, CG force fields are not completely flexible, but only include particular types of interactions between CG sites, e.g., nonbonded forces between pairs of sites. If the CG force field depends linearly on the force field parameters, then the vector valued functions that relate the CG forces to these parameters determine a set of basis vectors that span a vector subspace of CG force fields. The companion paper introduced a distance metric for the vector space of CG force fields and proved that the MS-CG variational principle determines the CG force force field that is within that vector subspace and that is closest to the force field determined by the many-body PMF. The present paper applies the MS-CG variational principle for parametrizing molecular CG force fields and derives a linear least squares problem for the parameter set determining the optimal approximation to this many-body PMF. Linear systems of equations for these CG force field parameters are derived and analyzed in terms of equilibrium structural correlation functions. Numerical calculations for a one-site CG model of methanol and a molecular CG model of the EMIM(+)NO(3) (-) ionic liquid are provided to illustrate the method.     

Modeling real dynamics in the coarse-grained representation of condensed phase systems

This work presents a systematic multiscale methodology to provide a more faithful representation of real dynamics in coarse-grained molecular simulation models. The theoretical formalism is based on the recently developed multiscale coarse-graining (MS-CG) method [S. Izvekov and G. A. Voth, J. Phys. Chem. B. 109, 2469 (2005); J. Chem. Phys. 123, 134105 (2005)] and relies on the generalized Langevin equation approach and its simpler Langevin equation limit. The friction coefficients are determined in multiscale fashion from the underlying all-atom molecular dynamics simulations using force-velocity and velocity-velocity correlation functions for the coarse-grained sites. The diffusion properties in the resulting CG Brownian dynamics simulations are shown to be quite accurate. The time dependence of the velocity autocorrelation function is also well-reproduced relative to the all-atom model if sufficient resolution of the CG sites is implemented.     

Mixed atomistic and coarse-grained molecular dynamics: simulation of a membrane-bound ion channel

The recently developed multiscale coarse-graining (MS-CG) method (Izvekov, S.; Voth, G. A. J. Phys. Chem. B 2005, 109, 2469; J. Chem. Phys. 2005, 123, 134105) is used to build a mixed all-atom and coarse-grained (AA-CG) model of the gramicidin A (gA) ion channel embedded in a dimyristoylphosphatidylcholine (DMPC) lipid bilayer and water environment. In this model, the gA peptide was described in full atomistic detail, while the lipid and water molecules were described using coarse-grained representations. The atom-CG and CG-CG interactions in the mixed AA-CG model were determined using the MS-CG method. Molecular dynamics (MD) simulations were performed using the resulting AA-CG model. The results from simulations of the AA-CG model compare very favorably to those from all-atom MD simulations of the entire system. Since the MS-CG method employs a general and systematic approach to obtain effective interactions from the underlying all-atom models, the present approach to rigorously develop mixed AA-CG models has the potential to be extended to many other systems.     

The multiscale coarse-graining method. IX. A general method for construction of three body coarse-grained force fields

The multiscale coarse-graining (MS-CG) method is a method for constructing a coarse-grained (CG) model of a system using data obtained from molecular dynamics simulations of the corresponding atomically detailed model. The formal statistical mechanical derivation of the method shows that the potential energy function extracted from an MS-CG calculation is a variational approximation for the true potential of mean force of the CG sites, one that becomes exact in the limit that a complete basis set is used in the variational calculation if enough data are obtained from the atomistic simulations. Most applications of the MS-CG method have employed a representation for the nonbonded part of the CG potential that is a sum of all possible pair interactions. This approach, despite being quite successful for some CG models, is inadequate for some others. Here we propose a systematic method for including three body terms as well as two body terms in the nonbonded part of the CG potential energy. The current method is more general than a previous version presented in a recent paper of this series [L. Larini, L. Lu, and G. A. Voth, J. Chem. Phys. 132, 164107 (2010)], in the sense that it does not make any restrictive choices for the functional form of the three body potential. We use hierarchical multiresolution functions that are similar to wavelets to develop very flexible basis function expansions with both two and three body basis functions. The variational problem is solved by a numerical technique that is capable of automatically selecting an appropriate subset of basis functions from a large initial set. We apply the method to two very different coarse-grained models: a solvent free model of a two component solution made of identical Lennard-Jones particles and a one site model of SPC/E water where a site is placed at the center of mass of each water molecule. These calculations show that the inclusion of three body terms in the nonbonded CG potential can lead to significant improvement in the accuracy of CG potentials and hence of CG simulations.     

Thermodynamic transferability of coarse-grained potentials for polymer-additive systems

In this work we study the transferability of systematically coarse-grained (CG) potentials for polymer-additive systems. The CG nonbonded potentials between the polymer (atactic polystyrene) and three different additives (ethylbenzene, methane and neopentane) are derived using the Conditional Reversible Work (CRW) method, recently proposed by us [Brini et al., Phys. Chem. Chem. Phys., 2011, 13, 10468-10474]. A CRW-based effective pair potential corresponds to the interaction free energy between the two atom groups of an atomistic parent model that represent the coarse-grained interaction sites. Since the CRW coarse-graining procedure does not involve any form of parameterisation, thermodynamic and structural properties of the condensed phase are predictions of the model. We show in this work that CRW-based CG models of polymer-additive systems are capable of predicting the correct structural correlations in the mixture. Furthermore, the excess chemical potentials of the additives obtained with the CRW-based CG models and the united-atom parent models are in satisfactory agreement and the CRW-based CG models show a good temperature transferability. The temperature transferability of the model is discussed by analysing the entropic and enthalpic contributions to the excess chemical potentials. We find that CRW-based CG models provide good predictions of the excess entropies, while discrepancies are observed in the excess enthalpies. Overall, we show that the CRW CG potentials are suitable to model structural and thermodynamic properties of polymer-penetrant systems.     

The multiscale coarse-graining method. VII. Free energy decomposition of coarse-grained effective potentials

The potential of mean force (PMF) with respect to coarse-grained (CG) coordinates is often calculated in order to study the molecular interactions in atomistic molecular dynamics (MD) simulations. The multiscale coarse-graining (MS-CG) approach enables the computation of the many-body PMF of an atomistic system in terms of the CG coordinates, which can be used to parameterize CG models based on all-atom configurations. We demonstrate here that the MS-CG method can also be used to analyze the CG interactions from atomistic MD trajectories via PMF calculations. In addition, MS-CG calculations at different temperatures are performed to decompose the PMF values into energetic and entropic contributions as a function of the CG coordinates, which provides more thermodynamic information regarding the atomistic system. Two numerical examples, liquid methanol and a dimyristoylphosphatidylcholine lipid bilayer, are presented. The results show that MS-CG can be used as an analysis tool, comparable to various free energy computation methods. The differences between the MS-CG approach and other PMF calculation methods, as well as the characteristics and advantages of MS-CG, are also discussed.     

Systematic coarse-graining of nanoparticle interactions in molecular dynamics simulation

A recently developed multiscale coarse-graining procedure [Izvekov, S.; Voth, G. A. J. Phys. Chem. B 2005, 109, 2469] is extended to derive coarse-grained models for nanoparticles. The methodology is applied to C(60) and to carbonaceous nanoparticles produced in combustion environments. The coarse-graining of the interparticle force field is accomplished applying a force-matching procedure to data obtained from trajectories and forces from all-atom MD simulations. The CG models are shown to reproduce accurately the structural properties of the nanoparticle systems studied, while allowing for MD simulations of much larger self-assembled nanoparticle systems.     

Markovian approximation in a coarse-grained description of atomic systems

The Markovian assumption stating that memory effects can be neglected is a crucial assumption in the theory of coarse-graining. We investigate the coarse-graining of a one-dimensional chain of oscillators where the atoms are grouped into clusters or blobs. When the interaction between oscillators is through Hookean springs, the cluster dynamics is non-Markovian, as has been recently noted by Cubero and Yaliraki [J. Chem. Phys. 122, 03418 (2005)]. When the oscillators interact through a nonlinear potential of the Lennard-Jones type, the dynamics turns out to be Markovian. The different behavior in both types of interactions is attributed to the persistence of sound waves in the harmonic case, which are strongly suppressed in the nonlinear case.     

Particle-based multiscale coarse graining with density-dependent potentials: application to molecular crystals (hexahydro-1,3,5-trinitro-s-triazine)

We describe the development of isotropic particle-based coarse-grain models for crystalline hexahydro-1,3,5-trinitro-s-triazine (RDX). The coarse graining employs the recently proposed multiscale coarse-graining (MS-CG) method, which is a particle-based force-matching approach for deriving free-energy effective interaction potentials. Though one-site and four-site coarse-grain (CG) models were parameterized from atomistic simulations of non-ordered (molten and ambient temperature amorphous) systems, the focus of the paper is a detailed study of the one-site model with a brief recourse to the four-site model. To improve the ability of the one-site model to be applied to crystalline phases at various pressures, it was found necessary to include explicit dependence on a particle density, and a new theory of local density-dependent MS-CG potentials is subsequently presented. The density-dependency is implemented through interpolation of MS-CG force fields derived at a preselected set of reference densities. The computationally economical procedure for obtaining the reference force fields starting from the interaction at ambient density is also described. The one-site MS-CG model adequately describes the atomistic lattice structure of α-RDX at ambient and high pressures, elastic and vibrational properties, pressure-volume curve up to P = 10 GPa, and the melting temperature. In the molten state, the model reproduces the correct pair structure at different pressures as well as higher order correlations. The potential of the MS-CG model is further evaluated in simulations of shocked crystalline RDX.     

Multiscale coarse-graining of ionic liquids

A recently developed multiscale coarse-graining (MS-CG) approach for obtaining coarse-grained force fields from fully atomistic molecular dynamics simulation is applied to the challenging case of the EMIM+NO3- ionic liquid. The force-matching in the MS-CG methodology is accomplished with an explicit separation of bonded and nonbonded forces. While the nonbonded forces are adopted from this force-matching approach, the bonded forces are obtained from fitting the statistical configurational data from the atomistic simulations. The many-body electronic polarizability is also successfully broken into effective pair interactions. With a virial constraint fixing the system pressure, the MS-CG models rebuild satisfactory structural and thermodynamic properties for different temperatures. The MS-CG model developed from a modest atomistic simulation is therefore suitable for simulating much larger systems, because the coarse-grained models show significant time integration efficiency. This approach is expected to be general for coarse-graining other ionic liquids, as well as many other liquid-state systems. The limitations of the present coarse-graining procedure are also discussed.     

The multiscale coarse-graining method. X. Improved algorithms for constructing coarse-grained potentials for molecular systems

The multiscale coarse-graining (MS-CG) method uses simulation data for an atomistic model of a system to construct a coarse-grained (CG) potential for a coarse-grained model of the system. The CG potential is a variational approximation for the true potential of mean force of the degrees of freedom retained in the CG model. The variational calculation uses information about the atomistic positions and forces in the simulation data. In principle, the resulting MS-CG potential will be an accurate representation of the true CG potential if the basis set for the variational calculation is complete enough and the canonical distribution of atomistic states is well sampled by the data set. In practice, atomistic configurations that have very high potential energy are not sampled. As a result there usually is a region of CG configuration space that is not sampled and about which the data set contains no information regarding the gradient of the true potential. The MS-CG potential obtained from a variational calculation will not necessarily be accurate in this unsampled region. A priori considerations make it clear that the true CG potential of mean force must be very large and positive in that region. To obtain an MS-CG potential whose behavior in the sampled region is determined by the atomistic data set, and whose behavior in the unsampled region is large and positive, it is necessary to intervene in the variational calculation in some way. In this paper, we discuss and compare two such methods of intervention, which have been used in previous MS-CG calculations for dealing with nonbonded interactions. For the test systems studied, the two methods give similar results and yield MS-CG potentials that are limited in accuracy only by the incompleteness of the basis set and the statistical error of associated with the set of atomistic configurations used. The use of such methods is important for obtaining accurate CG potentials.     

A multiscale coarse-graining method for biomolecular systems

A new approach is presented for obtaining coarse-grained (CG) force fields from fully atomistic molecular dynamics (MD) trajectories. The method is demonstrated by applying it to derive a CG model for the dimyristoylphosphatidylcholine (DMPC) lipid bilayer. The coarse-graining of the interparticle force field is accomplished by an application of a force-matching procedure to the force data obtained from an explicit atomistic MD simulation of the biomolecular system of interest. Hence, the method is termed a "multiscale" CG (MS-CG) approach in which explicit atomistic-level forces are propagated upward in scale to the coarse-grained level. The CG sites in the lipid bilayer application were associated with the centers-of-mass of atomic groups because of the simplicity in the evaluation of the forces acting on them from the atomistic data. The resulting CG lipid bilayer model is shown to accurately reproduce the structural properties of the phospholipid bilayer.     

Towards an understanding of many-particle effects in hydrophobic association in methane solutions

This paper applies the multiscale coarse-graining method [S. Izvekov and G. A. Voth, J. Phys. Chem. B 109, 2469 (2005); J. Chem. Phys. 123, 134105 (2005)] to analyze many-body effects in concentrated methane solutions. Pairwise decompositions of N-particle solute-solute potentials of mean force (PMFs), and the respective solvent cavity potentials, enthalpic, entropic, and heat capacity of hydrophobic association, are calculated directly from unconstrained molecular-dynamics simulations of methane solutions at different molar fractions, with the highest being 0.055. The many-body effects in hydrophobic hydration are further studied using N-methane PMFs, which are explicitly dependent on solvent coordinates.     

The multiscale coarse-graining method. VIII. Multiresolution hierarchical basis functions and basis function selection in the construction of coarse-grained force fields

The multiscale coarse-graining (MS-CG) method is a method for determining the effective potential energy function for a coarse-grained (CG) model of a molecular system using data obtained from molecular dynamics simulation of the corresponding atomically detailed model. The coarse-grained potential obtained using the MS-CG method is a variational approximation for the exact many-body potential of mean force for the coarse-grained sites. Here we propose a new numerical algorithm with noise suppression capabilities and enhanced numerical stability for the solution of the MS-CG variational problem. The new method, which is a variant of the elastic net method [Friedman et al., Ann. Appl. Stat. 1, 302 (2007)], allows us to construct a large basis set, and for each value of a so-called "penalty parameter" the method automatically chooses a subset of the basis that is most important for representing the MS-CG potential. The size of the subset increases as the penalty parameter is decreased. The appropriate value to choose for the penalty parameter is the one that gives a basis set that is large enough to fit the data in the simulation data set without fitting the noise. This procedure provides regularization to mitigate potential numerical problems in the associated linear least squares calculation, and it provides a way to avoid fitting statistical error. We also develop new basis functions that are similar to multiresolution Haar functions and that have the differentiability properties that are appropriate for representing CG potentials. We demonstrate the feasibility of the combined use of the elastic net method and the multiresolution basis functions by performing a variational calculation of the CG potential for a relatively simple system. We develop a method to choose the appropriate value of the penalty parameter to give the optimal basis set. The combined effect of the new basis functions and the regularization provided by the elastic net method opens the possibility of using very large basis sets for complicated CG systems with many interaction potentials without encountering numerical problems in the variational calculation.     

Local response dispersion method. II. Generalized multicenter interactions

Recently introduced local response dispersion method [T. Sato and H. Nakai, J. Chem. Phys. 131, 224104 (2009)], which is a first-principles alternative to empirical dispersion corrections in density functional theory, is implemented with generalized multicenter interactions involving both atomic and atomic pair polarizabilities. The generalization improves the asymptote of intermolecular interactions, reducing the mean absolute percentage error from about 30% to 6% in the molecular C(6) coefficients of more than 1000 dimers, compared to experimental values. The method is also applied to calculations of potential energy curves of molecules in the S22 database [P. Jure?ka et al., Phys. Chem. Chem. Phys. 8, 1985 (2006)]. The calculated potential energy curves are in a good agreement with reliable benchmarks recently published by Molnar et al. [J. Chem. Phys. 131, 065102 (2009)]. These improvements are achieved at the price of increasing complexity in the implementation, but without losing the computational efficiency of the previous two-center (atom-atom) formulation. A set of different truncations of two-center and three- or four-center interactions is shown to be optimal in the cost-performance balance.     

Multiscale coarse-graining of monosaccharides

A systematic multiscale coarse-graining (MS-CG) algorithm is applied to build coarse-grained models for monosaccharides in aqueous solution. The methodology is demonstrated for the example of alpha-D-glucopyranose. The nonbonded interactions are directly derived from the force-matching approach, whereas the bonded interactions are obtained through Boltzmann statistical analyses of the underlying atomistic trajectory. The MS-CG model is shown to reproduce many structural and thermodynamic properties in the constant NPT ensemble. Although the model is derived at a single temperature, pressure, and concentration, it is shown to be reasonably transferable to other thermodynamic states. In this model, long-range interactions are effectively mapped into short-range forces with a moderate cutoff and are evaluated by table look-up. As a result, molecular dynamics employing the MS-CG model is approximately 3 orders of magnitude more efficient than its atomistic counterpart. Consequently, the model is particularly suitable for simulating carbohydrate systems at large length and long time scales. Results for an alpha-(1-->4)-d-glucan with 14 glucose units are also presented, demonstrating that the MS-CG algorithm is also applicable to the coarse-graining of other saccharide systems.     

Homopairing possibilities of the DNA base thymine and the RNA base uracil: an ab initio density functional theory study

All planar homopairings of the DNA base thymine and the RNA base uracil are reported for the first time in this study. Using the idea of binding sites discussed in our previous work (Kelly et al. J. Phys. Chem. B 2005, 109, 11933; J. Phys. Chem. B 2005, 109, 22045) and ab initio density functional theory, we predict and relax 10 thymine and 10 uracil homopairs. The stabilization energies of the homopairs vary from just below zero to -0.82 eV. The results on the pair geometry and energetics are compared with those available in the literature. The collected data on all planar thymine and uracil homopairs can be used to construct the thymine and uracil superstructures seen experimentally on various surfaces.     

Multiscale coarse graining of liquid-state systems

A methodology is described to systematically derive coarse-grained (CG) force fields for molecular liquids from the underlying atomistic-scale forces. The coarse graining of an interparticle force field is accomplished by the application of a force-matching method to the trajectories and forces obtained from the atomistic trajectory and force data for the CG sites of the targeted system. The CG sites can be associated with the centers of mass of atomic groups because of the simplicity in the evaluation of forces acting on these sites from the atomistic data. The resulting system is called a multiscale coarse-grained (MS-CG) representation. The MS-CG method for liquids is applied here to water and methanol. For both liquids one-site and two-site CG representations without an explicit treatment of the long-ranged electrostatics have been derived. In addition, for water a two-site model having the explicit long-ranged electrostatics has been developed. To improve the thermodynamic properties (e.g., pressure and density) for the MS-CG models, the constraint for the instantaneous virial was included into the force-match procedure. The performance of the resulting models was evaluated against the underlying atomistic simulations and experiment. In contrast with existing approaches for coarse graining of liquid systems, the MS-CG approach is general, relies only on the interatomic interactions in the reference atomistic system.     

The structure and binding energies of the van der Waals complexes of Ar and N2 with phenol and its cation, studied by high level ab initio and density functional theory calculations

We have investigated, using both ab initio and density functional theory methods, the minimum energy structures and corresponding binding energies of the van der Waals complexes between phenol and argon or the nitrogen molecule, and the corresponding complexes involving the phenol cation. Structures were obtained at the MP2 level using a large basis, and the corresponding energies were corrected for basis set superposition error (BSSE), higher order electron correlation effects, and for basis set size. The structures of the global minima were further refined for the effects of BSSE and the corresponding binding energies were evaluated. For each neutral species, we find only a single true minimum, pi bonded for argon and OH bonded for nitrogen. For both cationic species, we find that the OH-bonded complex is preferred over other minima which we have identified as having Ar or N(2) between exogeneous atoms. The ab initio calculations are generally in excellent agreement with experimental binding energies and rotational constants. We find that the B3LYP functional is particularly poor at describing these complexes, while a density functional theory (DFT) method with an empirical correction for dispersive interactions (DFT-D) is very successful, as are some of the new functionals proposed by Zhao and Truhlar [J. Phys. Chem. A 109, 5656 (2005); J. Chem. Theory Comput. 2, 1009 (2006); Phys. Chem. Chem. Phys. 7, 2701 (2005); J. Phys. Chem. A 108, 6908 (2004)]. Both the ab initio and DFT-D methods accurately predict the intermolecular vibrational modes.