Neighbor List Artifacts in Molecular Dynamics Simulations

Hyuntae Kim1,2, Balázs Fábián1, Gerhard Hummer1,3

  • 1Department of Theoretical Biophysics, Max Planck Institute of Biophysics, Max-von-Laue Straße 3, 60438 Frankfurt am Main, Germany.

Insights

Molecular dynamics (MD) simulations using default GROMACS parameters can cause membranes to deform due to missed interactions. Adjusting cutoff radii and neighbor list updates prevents unphysical box deformations in biophysical simulations.

Area of Science:

  • Biophysical research
  • Computational chemistry
  • Molecular modeling

Background:

  • Molecular dynamics (MD) simulations are crucial in biophysics.
  • Default parameters in simulation packages like GROMACS are often used by nonexperts.
  • Previous studies have not fully addressed potential issues with default MD parameters.

Purpose of the Study:

  • To identify the cause of unphysical membrane deformation in GROMACS MD simulations using default parameters.
  • To investigate the impact of parameter choices on pressure tensor accuracy.
  • To provide guidelines for optimizing MD simulation parameters for accurate results.

Main Methods:

  • Performed MD simulations using the GROMACS package with default settings.
  • Analyzed membrane deformation and pressure tensor components.
  • Investigated the role of outer cutoff (r_l) and neighbor list update frequency (nstlist).
  • Tested simulations with neat water in both atomistic and coarse-grained models.

Main Results:

  • Default GROMACS parameters led to membrane crumpling due to missed nonbonded interactions from short cutoffs and infrequent updates.
  • Systematic pressure imbalances caused asymmetric box deformations.
  • Dual pair lists with dynamic pruning resulted in pressure oscillations.
  • Similar issues were observed in simulations of water.

Conclusions:

  • Default MD parameters, particularly short outer cutoffs and infrequent neighbor list updates, can introduce significant artifacts in simulations of large systems like membranes.
  • These artifacts, though minor in smaller systems, can become critical in coarse-grained simulations of large-scale biological systems.
  • Recommendations are provided for diagnosing and avoiding these issues by optimizing r_l and nstlist parameters.