*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dear Machius,
<-- snip -->
> I think one can indeed agree that the structure of an object does not
> change with the resolution one looks at it. From that, one can
> conclude, that one should use the same target values for bond
> lengths, etc., for all resolutions. <-- snip -->
As golden as this nugget of wisdom might look, it is basically a
misconception. If a crystal only gives low-resolution data, the molecule it
is made of does not have an "absolute" structure which can be thought as
being the same as if the crystal were diffracting to high resolution. There
is an intrinsic degree of disorder, i.e. of inequivalence between the
various copies of the molecule, and the low-resolution X-ray data will only
enable us to have access, at best, to an electron density map which is an
ensemble average of the densities for those inequivalent molecules. While
each copy of the molecule would have a geometry which obeys stereochemical
restraints quite well; but the smeared, ensemble-averaged density described
by the X-ray data cannot be accounted for by a single model which would
itself comply with those restraints to the same degree of accuracy.
One option is to model that disorder, e.g. by TLS - and there, it is
well known that this kind of disorder can interfere with ideal distances
(see Willis & Pryor). Another is to accommodate this kind of interference by
sticking naively to a single model (with B factors) but by "suitably"
relaxing the restraints in the disordered regions, to allow for the fact
that one is trying to account for the ensemble average of the electron
densities of many different local structures in terms of the electron
density for a single local structure.
Tadeusz's suggestion that a "suitable" scheme for doing this would
probably involve changing the weights of different kinds of restraints in a
specific resolution-dependent way (and not just through a common rescaling
factor) seems a good starting point. However one can expect that the final
answer will be more complicated: in the theory of TLS, it is not only the
variance of an apparent bond length (in the smeared density) which needs
resetting, but its expectation value as well (again, see Willis and Pryor).
With best wishes,
Gerard.
