R-CCS Cafe 第223回 第2部

Residue-level coarse-grained (CG) models have become one of the most prevalent tools in biomolecular simulations. To study large-scale biological phenomena in molecular dynamics (MD) simulations with CG models, unified processing of proteins and nucleic acids, as well as efficient parallel computations, are essential. We implement several CG models in the GENESIS MD software, covering potentials for both well-folded biomolecules and disordered regions. A single CG particle represents around ten heavy atoms in both proteins and nucleic acids in these models. We prepare the input data in CG MD simulations using GROMACS-style files generated from a newly developed toolbox, GENESIS-CG-tool. To optimize the performance in CG MD simulations, we employ multiple neighbor lists assigned to a different nonbonded interaction potential. We noticed that random number generations in the Langevin thermostat are bottlenecks in CG MD simulations. We then parallelized this procedure with MPI to advance the performance on PC clusters or supercomputers. We simulate the whole Herpes simplex virus capsid and large chromatin structures as examples of large-scale biomolecular simulations in GENESIS. We believe this framework extends accessible spatial and temporal scales of multi-scale simulations and enables studies of biological phenomena such as genome-scale chromatin folding or phase-separated membrane-less condensations. https://doi.org/10.1101/2021.10.21.465249