[gmx-users] reproducibility of PMF plot with two different starting structures

chris.neale at utoronto.ca chris.neale at utoronto.ca
Fri May 21 16:35:46 CEST 2010

Your run is not converged. You could (i) run more slowly, (ii) run  
lots of these non-equilibrium runs and treat them appropriately, or  
(iii) use an equilibrium method.

In any event, I suggest that you begin by (iv) repeat the run for both  
poses with different random seeds to see how much of the difference  
has nothing to do with the starting structure. If the different  
starting poses give you SMD results whose difference is no greater  
than the difference between poses, then you're probably just looking  
at statistical error between the two.

You asked if this error is ok... sorry, I can't judge that. But still  
you need to know what kind of error it is (statistical, or  
systematically based on starting pose). Note that you will always get  
some difference between two runs, your goal is to have a small enough  
difference that it doesn't change your conclusions.


-- original message --


I just want to know about the consistency we could obtain in plotting PMF.

I am using Gromacs 4.0.4 MPI version. Trying to find the PMF of the ligand
transport through the channel. I tried two SMDs (same pull rate, and force
constant) with different starting structures (only ligand poses and
position  are slightly  different). I performed the sampling of these two
pathways, the plots obtained are of  similar overall shapes, but peaks (max
, min) have some variations?

Could you please give your comment on this attempt? Is that expectable?

Thank you,


More information about the gromacs.org_gmx-users mailing list