Dear all,
First of all, congratulation on the new platform and thanks for all your efforts.
In a PMF simulation using WHAM, the windows are somehow that the reaction coordinate (RC = COM_group1 ----COM_group2) should have both positive and negative parts along Z direction, e.g COM_group1 locates in the middle of the reaction coordinate more or less. However, what the WHAM gives out is only the positive part of the RC, while all the .tpr and pullf.xvg files for the whole RC are available.
I wonder how I can have PMF along the whole reaction coordinate in this case?
Of course, one possible solution might be to run WHAM separately on windows of positive and negative parts of reaction coordinate and then merge the two profiles in a single profile later. But what I found out unexpectedly is that presence of each part (e.g negative part) in gmx WHAM, would affect on the PMF of the another part so that the PMF of the only positive part varies when the windows of the negative part are and are not considered.
No data point is found in the whole negative part of RC, when by using “-sym yes -min -7.2 -max 7.2” I enforce the gmx WHAM to go over the negative part of RC.
Do you also use -sym = yes in your original setup?
sym averages the pmf values around zero, linearly interpolating if between symmetric values around zero if the spacing does not match exactly.
cycl assumes that the coordinate you are looking at is periodic and that overall the PMF should connect at both ends. To ensure that, the whole profile is tilted to make sure that start- and endpoint have the same value.
cycl also assumes that you can let the profile start at 0, because you can periodically shift as you like.
Thanks.
Yes indeed, I used “-sym yes -min -7.2 -max 7.2” to enforce gmx WHAM to go over the negative part, but it could find no data point in the negative part.
It may relate to your setting in mdp file, can you share the Umbrella sampling setting part of your mdp file?
what is the output of grompp when you create tpr file in the “negative part” of your RC?
For example, if you set pull-coord1-geometry= distance, then grompp will tell your RC is alway positive
Thanks.
Please find below the US setting part.
The MD of each windows was performed using the pull-coord1-geometry = distance, and then if I use the TPR files from the windows no data could find for the negative part, I understand now that is because all the distance are positive.
Now, I just generated new TPR files using the pull-coord1-geometry = direction (no new simulation but I just used the pullf.xvg files from the previous simulation) and now I have both positive and negative parts of the reaction coordinate in PMF, please find attached figure for comparison.
There is now one problem: as attached figure shows the results of PMF are kind of different when different direction of -Z or +Z is chosen for the pull-coord1-vec = 0 0 -1.0 ;; 0 0 1.0. And also they differ with when the pull-coord1-geometry = distance is used in WHAM.
Could anybody kindly explain the reason for me, please? Which one should be the correct PMF?
I fed the WHAM with exactly similar input files both quality and quantity (except the pull-coord1-vec), however, some point around have no data for the +Z direction whereas that is not the case for the -Z direction.
Blue vs red line: in principle, they should be mirror symmetry by X-axis, but you used the result from one set of simulations for other sets up so I think that is the source of the problem why they are not symmetry.
Force in pullf.xvg is a vector quantity.
If you want to know why they are not similar, my suggestion is to run simulations again, using tpr and pullf.xvg that generate from the same setup, do not mix them together.
And also important thing is you have to make sure your simulations are converged.
the default value for pull-nstxout, pull-nstfout is fine, as I remember, it is 50 steps
Thanks Qvuvan,
I guess the blue and red PMF curves should have reflection symmetry respect to the (0, 0), not mirror symmetry by X-axis.
I wanted to avoid redoing the simulations.
