Multiscale Coarse-Graining of Polarizable Models through Force-Matched Dipole Fluctuations

Abstract

An extension of the multiscale coarse-graining method (MS-CG) to polarizable coarse-grain (CG) models is presented. In the extension, force matching is used to derive charged dimers that mimic the dipole behavior, including electronic polarizability, of fine resolution systems. The extended MS-CG method separates short-range and electrostatic forces and treats the polarization interactions by representing dipole fluctuations with a harmonic bond reminiscent of the Drude oscillator. The new method is first tested on several flexible alcohols, where the transferable head groups in the atomistic field lead to transferable electrostatics in the CG field. The method is then applied to a polarizable methanol model, where the CG dimer is able to match the atomistic dipole distribution. Force fields are benchmarked with radial distribution functions and dielectric constants. The static dielectric constants agree well with atomistic models, but the faster dynamics in the CG ensembles leads to significant deviations in the frequency-dependent permittivity. Model transferability is checked by measuring response to static external fields through a dipole coupling order parameter. Effective polarizabilities are inferred from the dimer harmonic bonds, and an ad hoc correction is used to improve one CG force field's dipole magnitude response to electric fields. Broader applicability is explored through a small set of amines, and several multibead models are tested where CG dipole orientations are restricted through intramolecular forces, capturing constructive and destructive dipole interactions within single molecules.

Extended coarse-grained dipole model for polar liquids: Application to bulk and confined water

“…The parameters are tuned such that with a loss of local structure, it reproduces the bulk water density, and diffusion at room temperature in a good agreement with experiments [25,26]. In a more systematic way, the force matching technique is used with the Drude-like model, to reproduce the radial distribution functions (RDFs) and the dielectric permittivity of some amines and alchohols such as methanol [27]. In this study, we systematically develop an extended dipole-based model for polar fluids that can accurately predict the radial distribution function (RDF), diffusion coefficient, and the bulk permittivity of the reference all-atom model.…”

Parametrizing hydrogen bond interactions in dissipative particle dynamics simulations: The case of water, methanol and their binary mixtures

“…3 the first peaks of both the atomistic and DPD simulations are quite similar to each other. However, the second and third peaks are slightly lower as well as shifted somewhat as compared to the MD simulations [54] employed by using a polarizable form of the TraPPE potential for methanol, signifying a longer-range order for the second and third shell interactions in the MD simulations. The set Morse potential parameters yield a good estimate of the first shell interactions, which are more concentrated than the remaining neighboring shells.…”

“…Fig.3. Radial Distribution Functions g(r) of the DPD simulations of pure methanol with the hydrogen bond interactions (red), classical DPD simulations that do not contain hydrogen bond attraction (blue), and the atomistic RDF (black)[54].…”

A generic force field for simulating native protein structures using dissipative particle dynamics

“…As mentioned earlier, we have adopted the Flory-Huggins parameters for the DPD beads of proteins from the work Truszkowski et al, 56 where molecular fragments were simulated in an aqueous solvent using the COMPASS force field. 85 The full interaction matrix of the repulsive force constant evaluated from eqn (8) is tabulated in Table 2, where P i s represent the different bead types as described in Fig. 2.…”

“…Consequently, there has been a rapid development of CG models and multiscale methods in the last few decades for simulating simple fluids, ionic liquids, biomembranes, proteins and transmembrane protein assemblies. [1][2][3][4][5][6][7][8][9][10][11] With an increase in virulent bacterial and viral infections, CG modelling of biological systems involving membraneprotein complexes has attracted considerable attention. [11][12][13][14] As a result, numerous CG protein models have been developed at various levels of reduced resolution ranging from one to five CG beads mapped to represent an amino acid residue.…”