[gmx-users] g_msd

[Xavier Periole]

Dear ram ram,
> I am trying to calculate the diffusion coefficient for my protein using
> g_msd. I issued the following command:
> 
> g_msd -f .xtc -s .pdb  -o .xvg
> 
> It gave me the diffusion coefficient as 0.068 (+/- 0.0279) * 10^-5 cm^2 s^
> -1  (for my whole system). The shape of the plot is not linear rather it is
> parabola (I think it might be due to the short period of simulation but I am
> not sure)
This is fine. The deviation from the linearity will appear in most cases
that differ from a simple water box.
> I also used g_analyze with the following commad:
> 
> g_analyze -f .xvg (the xvg from g_msd I have used here) -msd
> 
> This plot is linear but it considered only upto 5ns time period though my
> xtc/xvg file contains 10ns data. Could some one tell me why g_analyze
> considered only 5ns simulation time. Also how g_msd and g_analyze are
> different ?
g_msd uses/plots only half the simulation length by default. This is the
maximum "accepted" to be used for a msd curve. The reason is that after
that time range the statistics is really really bad.
Most of the time you should actually use much less than half of the run
to do your fit, don;t forget that your are assuming linearity and fit with
the Einstein equation. See the manual some details are given there I believe
or look for mean square displacement in google/wiki or any book.
> I learnt form one of Xavier Periole's postings, that g_msd is the
> instantaneous value of msd and g_analyze is the average over the entire
> simulation. 
I hope my post was a bit more clear than that! I probably meant that g_msd
gives the msd in function of time! I don't see what I meant with "g_analyse
being the average of the entire simulation"!
> The diffusion coefficient value I got
> is from g_msd, how can I calculate the value of the same by using g_analyze
> ?
Don't know! I may have known. Things are lost in translation.
> Also, while calculating diffusion coefficient, is it better to monitor a
> particular atom or a group of atoms or entire protein ?
This depends what you are looking for. To get the diffusion of a protien
you would want to use the center of mass of the protein.

XAvier.
> Thanks and Regards.

-----------------------------------------------------
XAvier Periole - PhD

Molecular Dynamics Group / NMR and Computation
University of Groningen
The Netherlands
-----------------------------------------------------