Abstract
Computational chemistry is an important tool in numerous scientific disciplines, including drug discovery, chemical reaction design, materials science, and structural biology. Coarse-grained models offer a simplified representation of molecular systems, that in theory enables efficient simulations of large-scale system. Over the last decade, there has been a considerable increase in the adoption of coarse-grained models for simulations of biomolecular systems. Therefore, critical and independent evaluation of such models is warranted. Here, the properties of crystals of amyloid peptides as well as organic molecules are evaluated using the Martini coarse-grained force field. We investigate whether this force field can accurately maintain the crystal structure of amyloid peptides and organic compounds and predict melting temperatures. The peptide crystals change shape in the simulations, in most cases drastically so. Radial distribution functions show that the distance between backbone beads representing intermolecular hydrogen bonds in β-sheets increases by about 1 ̊A in the simulation, breaking the crystals. In addition, the melting points of organic compounds are much lower in the Martini force field than in either an all-atom force field or experiment. Radial distribution functions for pyridine and phenol at 5 K show that the crystals transition into a glassy state when using Martini. Our results suggest that the Martini 3 model lacks the necessary level of specific interactions to accurately simulate peptide crystals or organic crystals without imposing artificial restraints. We speculate that the problems are exacerbated by the use of the 12-6 Lennard-Jones potential and suggest that a different, softer, potential could prove advantageous in harnessing this model for crystal simulations.
Supplementary weblinks
Title
MD Benchmarking data and inputs
Description
Input files to evaluate properties of organic solids using molecular simulation. For the simulations we use GROMACS and some Python scripts in this repository. The scripts are adapted to some of the CPU clusters we have access to, so will need to be adapted by anyone trying to reproduce parts of this work.
Actions
View