Hi All I am running a steered MD simulation and using pull coordinate and pull force output after every 500 steps and the total simulation steps is 50000; However, due to a fast transition from during the ends of the simulation, in the last 100 to 150 ps simulation time the pull force file has not updated. I can get the pull coordinates using gmx distance utility. I need to plot pull force vs time profile and also the maximum force required for the process. I have observed that the pull-coordinate (pullx) and pull force (pullf) files do not get updated like other output files in gromacs, there is a very noticable lag. It looks like although the simulation has stopped but the pullx and pullf files were not updated properly, so some values are missing. Is there any tool in gromacs that I can use to retrieve complete pull force time profile for the entire simulation run time?
Output files all always buffered. Since the pull output files usually contain little output, they are flushed to disc less frequently. You mdrun process must have been killed abruptly, as mdrun tries to flush all buffer before exiting. If you don’t have output in both pull files there is nothing you can do, except for restarting from the last checkpoint file, if that is from before where you are missing data, which I guess it is.
Thank you prof. Hess for your reply. In my case since I am trying to pull a ligand through a membrane protein and running a NPT steered MD, at a certain point of time the reaction coordinate distance becomes greater than half box length once the ligand has been moved out of the protein in the solution. Then, the simulation stops abruptly as you have rightly pointed out. That is probably the reason why the pullx and pullf files are not updated. So, if I restart the simulation again from the checkpoint file, I guess a similar thing is going to happen and the simulation may get aborted abruptly before the pullx and pullf files are updated. As for pullx file, I can still get the values using gmx distance. Is there any possibility to retrieve the pullf file for such cases? Currently I am using pull-coord-geometry=direction; Is direction-periodic a good option to use as pull-coord-geometry instead of only direction? However as per gromacs manual direction-periodic use is recommended for non-dynamic simulation box and my simulation is NPT ensemble.
I would think mdrun should flush the before exiting with a fatal error. Are you running with MPI?
But the solution is simply to compute how many steps you can take before the distance reaches half the box size.
Yes I am running mdrun with MPI, and also I have checked that after approximately 100000 steps or 200 ps is the time-duration between the last update of the pullx file and the last frame of the simulation. So, are you suggesting that I should use that many steps in a new mdp file (that is make a new mdp file with nsteps = 100000) and run grompp with last updated checkpoint file and then run a mdrun; I have done something similar hoping that pullx, pullf files will be updated in time, but I get empty pullx and pullf files.
If you have suggested otherwise could you please elaborate on how do you think we can retrieve the full time-profile of the pullx and pullf files