On Mon, Sep 17, 2012 at 11:12 AM, Yuri
However, I have had a couple of cases where anisotropic B-factor refinement significantly improved R-work and R-free, while maintaining a reasonable R-gap, for lower resolution models (1.4-1.5 A, around 70,000 reflections). What is the proper way of modelling the B-factors?
I have also seen the anisotropic refinement work very well at this resolution, but there are certainly exceptions. Pavel has recommended that you leave the waters isotropic until about 1.2A, however (there are shortcuts for setting the selections for this and other common scenarios in the GUI). At any rate, if you really want to be certain, don't think there's any substitute for running the refinement both ways, isotropic and anisotropic, and seeing which works best. There are probably different opinions on how to judge whether anisotropic refinement is justified, but R-free should drop by at least 0.5%, and 1% is probably a more rigorous cutoff. One of the recent PDB-REDO papers had a description of their decision-making process for this, which I thought seemed sensible (they're using REFMAC, not Phenix, but the principle is the same in either case). -Nat