[gmx-users] Problem with Gromacs and Charmm force field dump rates

[Peter Kroon]

Hi Vivien,


My answers are inline


On 08-02-18 12:28, Vivien WALTER wrote:
> Dear all,
>
> We are experiencing a strange problem with all-atom simulations using
> Gromacs and Charmm force field, and we have trouble to sort it out.
>
> We are simulating a single-ring sugar and we observe radically different
> behaviors of the molecule as a function of the simulation dump rate.
> Starting from the same .gro file with continuation = yes (so using the same
> initial velocities written in the file), the interactions of the sugar with
> its surroundings is different between two simulations if the following set
> of parameters are changed: nstxout, nstvout, nstenergy and nstlog. Between
> the two .mdp files, absolutely no other parameters are changed. As an
> example of what differs between the simulation, the box size along X and Y
> axis are increasing with low dump rates (2500 steps) but decreases with
> high dump rates (25000 steps).
Increase/decrease by how much? Is it converged?
>
> We tried to reproduce the problem with a box filled with only water
> molecules, but in this configuration there is no difference between our
> results when different dump rates are used. Following the previous example,
> the evolutions of the box size superimpose perfectly for different dump
> rate.
> We tried to change the conformation of the sugar to see if it would have an
> impact on the problem observed, but the problem remains.
>
> The forcefield used is Charmm36 (version from june 2015) and Gromacs
> version is 2016.4. For the sugar, the structure file was constructed
> according to the more stable chair conformation of the saccharide, but the
> force field of the molecule was directly taken from the standard Charmm
> force field parameters. Water model used is TIP3P.
> For the simulation parameters, the pressure coupling used is a
> semi-isotropic Parrinello-Rahman set at 1.0 bar in both directions
> (compressibility 4.5e-5 and tau_p equals to 2.0). For the temperature
> coupling, we used first V-Rescale, but as we suspected the stochastic
> process to lead to different results for different dump rates, we also
> tried a Nose-Hoover thermostat. In both cases, the three molecular groups
> are coupled separately at the same temperature and the same tau_t (0.4)
Don't do this (http://www.gromacs.org/Documentation/Terminology/Thermostats)
> The rest of the main options for the .mdp file are taken from the
> recommanded parameters for using Charmm36 in Gromacs (
> http://www.gromacs.org/Documentation/Terminology/Force_Fields/CHARMM )
>
> Is it a known problem of Gromacs using Charmm36? Is there a way to fix the
> problem and obtain identical results, or at least similar evolutions?
Simulations perfomed by Gromacs are not binary reproducible, so you will
(never) get exactly the same result twice starting from the same point.
This is inherent in how the simulation is performed.
http://www.gromacs.org/Documentation/Terminology/Reproducibility
This is why sampling is so important, to show that the fluctuations you
see average out and are not depending on stochastic errors or the
starting conformation.
>
> Thanks you in advance for your help,
> Best regards,
> Vivien WALTER
>
Peter