I am aware of to opposing schools of thought about modelling residues
without density:
- You can argue that since you know those residues are present in the
sequence, you can model them, even though there is no or hardly any
density. If you do so, you must set the occupancy of those residues to
zero. You might model according to the Ramachandran plot but
nevertheless will probably be producing some fictious stretch in the
final model
- On the other hand the model you produce is most likely going to be
published one way or the other, i.e., other people are going to look at
it. While Coot marks atoms with zero occupancy, many other PDB viewers
do not, and many end users may not have a crystallographic background.
So they may not be aware of the meaning of 'zero occupancy' and take
whatever you built for granted. In order to avoid end users drawing
conclusions from something which was not actually measured, you should
not model anything for which you don't see any density. Generally I
would not even model something if the density is only visible at a
sigma level below 1!
I strongly support the second school since my idea of crystallography is
not to produce something pretty but something useful. However, both
notions have their arguments, of course.
Tim
--
Tim Gruene
Institut fuer anorganische Chemie
Tammannstr. 4
D-37077 Goettingen
GPG Key ID = A46BEE1A
On Mon, 8 Dec 2008, Andy Millston wrote:
> I am trying to build a model of a 60000 Da protein from the diffraction
> data collected at 2.0 A resolution. There is a 10-residue stretch that
> has such bad electron density that even at 0.4 sigma level one can
> hardly see any well defined density for residues with long side chains.
>
> My question is: are such poorly defined regions left unmodeled in
> protein structures? Or is it conventional to model the whole chain no
> matter how poor the density? The region in question is in the middle of
> the chain and has several long side chain residues - both charged and
> uncharged.
>
> AM
>
>
>
|