[gmx-users] Average Structure outputted from g_rmsf

[Mark Abraham]

Arneh Babakhani wrote:
> Hi,
> 
> When outputting the average structure from g_rmsf using the -ox option,
> some of my residues in that average structure pdb have awkward coordinates
> (when you visualize it in VMD, it looks very strange).  Sometimes atoms
> are placed very close to each other, with really bad bond angles.  I
> suppose this is a facet of the averaging scheme in g_rmsf.

Yes. If you ask for an not-necessarily-physical result, you usually get 
one... see http://wiki.gromacs.org/index.php/Average_Structure

> Is there a way to correct for this, and to obtain an averagestructure.pdb
> that looks more reasonable coordinates and bond angles? (I ask b/c I'd
> like to use this averagestructure.pdb in further studies, such as
> docking.)

"Correcting for" is not quite the right phrasing here... :-) It's 
already correct according to the algorithm you used... it's not the 
algorithm's fault you actually wanted a physical structure. One approach 
you can use is to select the structure with the smallest RMSD from the 
average, kind of like selecting the median as the central measure rather 
than the mean of a numerical data set. Or you can take the nonphysical 
structure and do EM on it... One needs to apply brain, of course!

Mark