On Fri, Feb 1, 2013 at 5:31 AM, Urszula Uciechowska <
urszula.uciechowska.chem.umu.se> wrote:
> Dear Amber users,
>
> I am trying to get an average structure, using the the following script in
> Ptraj:
>
> trajin PROD6.mdcrd
> center :1-445 mass origin
> image origin center
> rms first mass out rms.dat :1-445
> average model-1_average.pdb pdb
>
> But in the average structure I got strange connections between oxygen
> atoms, please see attachment.
>
>
> Could anyone help me to figure out what went wrong with the calculation?
> How should I modify the script?
>
Nothing went wrong. Depending on the program you use for visualization,
bonds are drawn according to a distance criteria, so those bonds aren't
actually 'there'. To get sensible bonds drawn, you should load the
topology file and then the PDB file as a frame on top of that (VMD can do
this). That should display only the bonds present in the topology file.
Another misconception is the physical significance of an average structure.
An average structure is chemically ridiculous. For instance, if you have
a freely-rotating methyl group, then all three hydrogen atoms attached to
that methyl group are rotationally degenerate, and will therefore have the
SAME position when averaging a converged trajectory. Also, these hydrogen
atoms travel in a 'circle' during the rotation. The average position of
all points in a circle is the center of that circle. So not only will all
three hydrogens appear to be on top of one another, they will also be
almost on top of the carbon.
Your system appears to have multiple carbonyl groups, in which the carbonyl
oxygens may be similarly degenerate (or close to it). Therefore, their
average position may be the same if freely rotating. Or they will be close
if the rotation is not free or the simulation not converged. This doesn't
make your average structure 'wrong.'
The average structure is very useful for some things (for instance, you
should calculate RMS fluctuations and B-factors with respect to an average
reference), but it is sometimes unexpectedly useless for other things: like
using as a starting point for molecular dynamics. For many of the tasks
where you _think_ you want an average structure, a better choice would be
picking a representative structure of the highest-populated cluster when
doing cluster analysis.
Hopefully this was helpful,
Jason
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Feb 01 2013 - 05:30:04 PST
