David Briggs wrote:
I always leave these atoms in, in a preferred rotamer (with
1.0 occupancy!) and let the B-values refine to >100. These high B-values
clearly indicate the issue at hand, and you can colour the model
accordingly.
...assuming the end-user knows what a B-factor is...
If you can't say with certainty that it's there... why detract from
the accuracy of your model by including stuff that isn't in the data.
you are artificially increasing the R-factor of your model by making
your obs:param ratio less favourable.
I personally exclude atoms from the model that are not visible in the
electron density.... with the caveat that I use my judgment (as we all
do) when it comes to "visible".
However, if the B-factor is ignored by end-users, then leaving out side
chains that have a large degree of mobility in the structure will not
make end-user's calculations any more accurate (and will be less
accurate if they model an interaction with a beta-carbon-truncated LYS
in place of an entire LYS). It would obviously help if the software
gives warnings about unusual features in the PBD for novice users.
I propose that we adopt a new value for PDB files :) The D-value
(density) which is a measure of the <D/sigD> for each atom. Thus, when
a PBD file is deposited, depositors can be asked to justify/explain the
reason for including atoms with values less than a cutoff (0.5?) with
an option to automatically remove them and re-refine the structure with
the atoms removed.
Seriously though, if this (simple?) feature were to be built-in to
refinement software I think that it would be easier for a large
community to reach an agreement on how to represent disorder in atomic
models.
--
Jason Greenwald
UPR9050 CNRS
Pole API - ESBS
Bd Sebastien Brandt
BP 10413
67412 Illkirch Cedex
France
Tel 33 (0)3 90 24 47 31
Fax 33 (0)3 90 24 48 29
[EMAIL PROTECTED]
http://recepteurs.u-strasbg.fr