[gmx-users] Position Restraints MD

[Mark Abraham]

Philip Shushkov wrote:
> Dear GMX-USERS,
> 
> I would like to ask you for some advice because I have serious problem 
> in setting up a PR MD in GROMACS of DPPC monolayer. Prior to actual MD 
> simulation I have done exhaustive energy minimization up to 10 kJ/mol*nm 
> convergence criterion. Firstly, I minimized the system with restraints 
> applied to the heavy atoms of the lipid molecules and then the whole 
> system consisting of 81 DPPC molecules and 5575 water molecules. 
> Afterward, in order to avoid some unexpected behavior I minimized the 
> system with less stringent restraints on the heavy lipid atom which I 
> would keep in the course of the equilibration, a final step before the 
> equilibrating MD run. However, though the three stage minimization 
> procedure which surely relaxed all bad contacts, the first step of the 
> MD integration gave high positive energy due to the position restraints 
> contribution. Some trial and error steps allowed me to find that the 
> problem was in the grompp program and may be the parallelization of 
> GROMACS. The position restraints were OK with *.tpr file constructed 
> without the -shuffle option of grompp. Unfortunately, now the LINCS 
> algorithm crashes and obviously the systems tends to explode. But if I 
> again turn off the restraints and prepare the *.tpr with -shuffle using 
> the same starting configuration as before, everything looks very well. 
> In conclusion, I would like to humbly ask you where the problem could be 
> and are there any problems reported concerning PR constraint molecular 
> dynamics in parallel mode in GROMACS.

Position restraints and analytical constraints on the same atoms are 
likely to lead to numerical problems, particularly if your initial 
structure (to which you are presumably restraining) doesn't satisfy the 
constraints (which get re-inforced every integration step).

You can eliminate the parallel code as a source of problems by doing a 
test run on one processor.

Mark