Hi Everyone,
I am sorry to be such a pain about this but today I realized where
probably most of the problems with distance calculations and therefore
NOE-violations, distance based assignments and so forth originate.
There are three different possibilities of assignment for any prochiral
pair:
1) I have true stereochemical assignments (a state that is hard to
achieve nad probably not worth going for in the most cases)
2) I have to shifts for a given pair, but I don't know which shift
belongs to which atom (most common case)
3) The shifts are degenerate and I cannot distinguish them.
for the caculation of distances Analysis always assumes 3) to be the
case (except for the calculation of the violation list) and refers to
this as struct Distance which is the 'NOE sum' over all atoms of any
involved stereo pair of a given NOE. This is fine when I first try to
figure out which NOEs are possible based on a homology model or the
structure of another state of my protein when this has correct
stereochemistry.
However, during the the course of the structure calculation, one usually
ends up with something different. For every prochiral pair that falls
into 2) the stereo unspecific assignments eg Hba and Hbb are converted
into HB2 and HB3 based on a fixed rule. Therefore the structures that
originate from these restraints may have wrong stereochemistry of
prochirals (due to swapping during the calculation) but by that virtue
the unspecific prochiral assignments are converted into something that
is much closer to case 1) than it is to case 3).
Therefore I think it would be sensible to somehow flag the structures on
import whether they have correct stereochemistry that might be
incompatible with the fixed mapping of Hba and Hbb or have been subject
to swapping to accompany for that fixed mapping. In the later case it
would be far more sensible (and transparent to the user) to treat the
distances as if one had true stereochemical assignments (if that is
implemented yet) since this would reflect the look of the restraints to
the structure calculation software much closer than the current way.
Especially since this would be in the assignment dialogue and the Link
NOE popup it would provide a far more sensible way to look for
alternative assignments for violated restraints as the current state
where a peak that is in clear violation for eg. Xplor as well as the
calulate violations function within Analysis might be in perfect
agreement with the 'struct distance' as it is calculated at the moment.
One day, when SAIL has come upon us everything will fall under case 1)
anyway ;) ...
Hoping to be not too annoying,
Christoph
--
---------------------------------------------------------
Dr. Christoph Brockmann
Division of Structural Studies
MRC Laboratory of Molecular Biology
Hills Road
Cambridge, CB2 2QH
phone: +44-1223-40-2261 or +44-1223-40-2068
e-mail: [log in to unmask]
---------------------------------------------------------
|