• ## Outreach

From: Axel Kohlmeyer (akohlmey_at_cmm.chem.upenn.edu)
Date: Tue Oct 21 2008 - 15:53:07 CDT

On Tue, 21 Oct 2008, Tomek Wlodarski wrote:

TW> Dear Axel,
TW>
TW> Frankly I do not know yet how this averaging works, but do you think
TW> this approach is physical? I mean after averaging do I get structure

in my opinion, it is as physical as averaging over the density maps
that you create with volmaptool. all you do is replacing the coordinates
with a spherical density gaussian that has always the same shape (if
you'd do an electron structure calculation, that would be different)
for each atom at each step. the result that you get should be equivalent
of computing a density map for the average structure that is smoothened
by something that is proportional to the RSMD of the fit for the
individual atoms.

TW> which could be present in distribution of conformations for this
TW> protein?

very likely not. but the averaging over density maps has the same
problem. consider a simple example: a methyl group stuck on a molecule
that is otherwise fixed in space. if the methyl group rotates freely,
the the C-H bonds in the averaged structure will collapse. if you
compute the density map the H atom density will end up as smooth
density "ring" around the C, you cannot fit a CH3 group into this
so well (since the density is much lower) and particularly you
cannot identify the orientation. so basically you end up with
the same information. the RMSD tells you how much each structure
will fit to the average and the RMSF tells you how reliable the
average for each atom is. you can try to improve the fit by picking
some atom property (mass or other) as weighting parameter.

the RMSD fit will fail, when you want to compare two very different
structures that have a similar shape. this is where you'd need
a density (i.e. shape) based approach.

cheers,
axel.

TW> Thanks for replay,
TW> Best
TW>
TW> Tomek
TW>

```--
=======================================================================
Axel Kohlmeyer   akohlmey_at_cmm.chem.upenn.edu   http://www.cmm.upenn.edu
Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.
```