Has anyone experienced differences in B-factor refinement depending on
whether Xplor v3.1 or Xplor v3.851 is used?
I originally used Xplor v3.851 to refine a protein model against a 2.0 A
data set. It has Rfree=24.5%, Rcrys=21.5%, when bulk-solvent modelling
is included. The model includes two substrate molecules which, with
occupancies set to 0.5, have 4 of the 22 atoms with B=100.
I had some problems with the Fo-Fc map (which have since been sorted
out) so I switched to v3.1 and found that the B-factors for these
substrates refined to ~30-40 (occupancies set to 1.0). After many round
of model building and refinement in v3.1, I now have another model with
If I refine this model in v3.851, the B-factors for the substrates go
back up to 90-100 (and the R-factors increase to Rfree=25.5%,
I use the brefinement.inp files recommended for each Xplor version and
the parameter files specified within these correspond to those for
v3.851. The weights, resolution range, sigma cutoff, etc are the same
The Fo-Fc map produced for the v3.851 refined model is much noisier,
with a large negative peak over the substrate molecules. This is
presumably because v3.851 does not refine B-factors to values above 100.
However, the 2Fo-Fc map for this model is slightly better, particularly
for two other ligands bound to the protein.
Has anyone experienced similar problems with B-factor refinement using
the two versions of Xplor? Does anyone have any suggestions as to how
to solve these differences in B-factor refinement?
Centre for Drug Design and Development
University of Queensland
vos at biosci.uq.oz.au