[gmx-users] RMSD calculations by g_rms

[Tsjerk Wassenaar]

Hi Monika, Xavier,

You can't really judge a run from the rmsd ;)

g_rms and VMD should the exact same result. You might check the reference
> frame you use and make sure you fit your protein using the same set of
> atoms.
>

I concur, and Xavier, you hit the spot with referring to the reference
frame.

> 2. I have given two consecutive 3ns runs in continuation. first for 3ns,
> > then next again for 3ns. In latter, I used unconstrained-start=yes. When
> > I am concatenating these xtc files, to get total of 6ns, and then doing
> > g_rms, I am getting a dip at 3000 and there it seems to follow earlier
> > pattern of VMD.
> > that 1 to 3000ps- values of rmsd decreases from 0.5 to 0.005 nm,
> > then 3001 to 6000ps - values of rmsd increases from 0.0698 to 0.248 nm.
>
> Check again your reference frame, and you may have made a mistake in your
> continuation run, I mean the way you actually continued it.
>

So, what happens when you get a dip in the rmsd? The structure comes close
to your reference structure. Or, the other way around, your reference
structure seems to match the structure at the break of the trajectories.
Thus, it seems that you have used the continuation run input file as the
reference frame. I think that VMD only extracts the topology from the .tpr
file and does the rmsd calculation with respect to the first set of
coordinates in the trajectory.


> > In case of alone rmsd calculations, 3ns rmsd showed increasing
> > deviations from 0.0005 to 0.509 nm.
> > 6ns run rmsd too showed increasing deviations from 0.00 to 0.248nm
>
> This does not look like a regular continuation run. But again you
> reference frame might be the problem.
>

Well, the rmsd tends to rise more during the first part of the simulation
(relaxation effects, etc). So if you then take the run input file from the
continuation run and use it to calculate the rmsd of the second part, it's
likely that you'll see less of a rise in the rmsd.


> > But when I am concatenating them, then the 3ns run showed deviations
> > down (different from its original) and 6ns showed up (similar to its
> > original).
>
> Then the reference frame is not the problem but the way you continued.
> Do you fit your protein before the rmsd calculation?
>

Au contraire, the reference is the problem (or actually cause).

>
> > I really dont know what is happening and why?? Where the things are
> > going hay-ward??
> >
>

Because of an inconsistent approach. Please think carefully about which
reference file you use where. In these cases it's best to have a
start.tpror something, which corresponds to the initial starting
structure of your
simulation, for analysis.

I hope it helps and best wishes for 2008 ;)

Tsjerk

-- 
Tsjerk A. Wassenaar, Ph.D.
Junior UD (post-doc)
Biomolecular NMR, Bijvoet Center
Utrecht University
Padualaan 8
3584 CH Utrecht
The Netherlands
P: +31-30-2539931
F: +31-30-2537623
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20080101/295cbbe8/attachment.html>