[gmx-users] Autocorrelation times from gmx wham and analyze

[Gmx QA]

Hi all,

I am running a series of umbrella sampling simulations of a drug across a
membrane, with the final aim to determine permeability using the local
diffusion model.

I have come across papers where the (necessary) autocorrelation times of
the mean-squared fluctuation of the coordinates are calculated using gmx
wham, and this is also what I have done like this:

gmx wham -if pullf-files.dat -it tpr-files.dat -o -hist -ac -oiact


 and then the ac-times are written to the file iact.xvg.


To understand better the procedure however, I also wanted to do a similar
calculation manually.

I therefore did this:


gmx traj -f ../umbrella0.part0001.xtc -s ../umbrella0.tpr -ox
umbrella0_z_com.xvg -com -nox -noy


 to extract the z-com-coordinates of the drug molecule for a particular us
window.


Then:


gmx analyze -f umbrella0_z_com.xvg -ac -fitfn exp


Followed by:


gmx analyze -f autocorr.xvg  -integrate


to get what I though was going to be the equivalent acf-time tau. However,
for this umbrella window gmx wham gives me tau =  2.25 whereas the second
approach gives a value of 122. What could be the reason for this
difference? There is some info in cmx wham that says that the act
calculation is cut at a value of 0.05, and that is something that I haven't
been able to replicate in my second approach, but any insights are highly
appreciated.


Cheers

/PK