VMD-L Mailing List

From: Robin Bush (rmbush_at_uci.edu)
Date: Wed Sep 14 2005 - 07:50:56 CDT

Thanks Dan. You may have wondered why I was doing this in the first
place.... I just started using VMD to study very similar structures,
and the VMD per-residue RMSD plots didn't look like what I expected. I
wondered what the program would do with two identical files. The
returned RMSD values showed some pattern, being higher at the ends,
with some oscillations in the middle the protein. The oscillations
sort of coincided with gross structural features. This noise is tiny
with respect to the real distances between my structures, so I don't
care that it is there. But I did wonder where it came from!

Robin

On Sep 13, 2005, at 7:30 AM, Dan Wright wrote:

> Robin,
>
> My understanding of the operation of STAMP (the program used to
> perform the alignment) is as follows:
>
> 1) rapidly scan the (2 or more) molecules to be aligned to fit them to
> each other roughly
> 2) progressively refine the alignment using an iterative dynamic
> programming approach
>
> In step (1), the molecules to be aligned have their coordinates
> shifted around to find a starting position for performing step 2. Of
> course, STAMP would not need to do this in the case of identical
> molecules. Hoever, since it does not know that the molecules are
> identical before starting (and does not check for identical molecules
> because of course it is assumed that the user wants to align
> structures that differ in some respect), it still performs step 1
> which, in the case of identical structures, has the effect of
> un-aligning them. Step 2 then optimizes the rough alignment, but can
> never converge on the 100% "perfect" solution that existed before
> running STAMP in the first place. You may be able to get closer by
> messing with the STAMP settings before running it.
>
> In short, STAMP was not designed for this case; it is designed to
> align homologous but non-identical structures. If you do need to
> align identical structures, using VMD's built-in "measure fit" command
> would be better. It will fit atoms from 2 structures to each other
> based solely on RMSD.
>
> Dan
>
> Robin Bush wrote:
>> Hi,
>> If you load identical copies of a molecule, then do a structural
>> alignment and calculate the RMSD per residue, the resulting values,
>> while small, are not all zero.
>> Can anyone tell me why?
>> All I did was copy a pdb file, then load both the original and the
>> copy, and click away as follows...
>> VMD Main -> Extensions -> Analysis -> Multiple Alignment
>> Multiple Alignment -> Alignment -> Run Structural Alignment
>> Multiple Alignment -> Tools -> RMSD Tools -> RMSD Per Residue
>> Thanks,
>> Robin
>
> --
>
> Dan Wright
> (dtwright_at_uiuc.edu)
> (http://www.scs.uiuc.edu/)
> (UNIX Systems Administrator, School of Chemical Sciences, UIUC)
> (333-1728)
>
____________________________________

Dr. Robin Bush
Dept. of Ecology and Evolutionary Biology
359 Steinhaus
University of California
Irvine, CA 92697

rmbush_at_uci.edu
phone: 949-824-2243
fax: 949-824-2181
____________________________________