[gmx-users] rerun confusion

Mark Abraham Mark.Abraham at anu.edu.au
Fri Jul 1 01:45:54 CEST 2011

On 1/07/2011 1:57 AM, SebastianWaltz wrote:
> Hallo all together,
> I am working on a system of a small peptide solvated in CHCl3.
> I want to obtain the forces on the peptide atoms  induced by the
> solvent. For this I used the rerun option of mdrun_d. I did it now in
> two ways:
> 1st: using trjconv to get the pure peptide trajectory. When using the
> rerun option with no special configurations of the .mdp file. If the
> obtained force are pair additive I should get the wanted result if I
> subtract from the forces of the original trajectory the rerun forces.

Not necessarily. Reruns do neighbour searching every frame, which you 
may not have been doing in the original trajectory. Obviously, you'll 
need to omit all the solvent atoms after the subtraction.

> 2nd: using the original trajectory and changing the options of the .mdp
> file as follows:
> energygrps          =  Protein SOL
> energygrp_excl        =  Protein Protein SOL SOL
> and changing the coulomb type to Cut-off because of the separation
> problem using PME.
> Comparing this two results I realize that they are totally different.
> There is no correlation to detect between this two.

If the first used PME and rcoulomb is the same, then this is expected.

> I am wondering now which way to go to get the correct answer for my
> question. I probably just do not understand the energygrps and
> energygrp_excl option right.

The way you are using them above is correct. However, your two 
approaches should give the same results if you use the same 
electrostatics regime for both reruns.

> Using rerun with the same trajectory and the same .mdp option except
> changing PME to Cut-off gives the same result as the original trajectory
> or at least up to the last 3 decimal places of a double value.

That's inconceivable for well-formed PME. I think you may be mismatching 


> I would be really happy if someone can help me
> Thanks a lot
> Sebastian

More information about the gromacs.org_gmx-users mailing list