[phenixbb] why are my B factors being "refined"

[Pavel Afonine]

Hi Scott,

yes, I just reproduced it using a fake one-atom model and generated Fobs...

The output B-factors are different because of anisotropic scaling: the 
trace of overall scale matrix is subtracted from it and is added to Bsol 
and all atomic B-factors. This is described in details in:

http://www.phenix-online.org/newsletter/

see "On atomic displacement..." article.

I agree, strictly speaking you are right: you did not ask phenix.refine 
to do anything with atomic B-factors, so they must be the same in input 
and output files. I guess I make this sure in most of other case, but I 
most likely overlooked rigid-body refinement case (it has its own 
bulk-solvent correction and scaling step).

I just made this change, so B-factors will stay the same in your case. 
It will be available in a next build (in a day or two).

Pavel.

> Hello Phenix community,
>
> I'm currently using nightly build 718, but this behavior is exhibited since sometime in the 1.6.xx releases too.
>
> I've turned off ADP refinement (only doing rigid_body refinement) and set the b_iso specifically as such:
>
>   modify_start_model {
>      selection = None
>      adp {
>        atom_selection = None
>        randomize = None
>        set_b_iso = 55
>        convert_to_isotropic = None
>        convert_to_anisotropic = None
>        shift_b_iso = None
>        scale_adp = None
>      }
>
>
> Yet the output PDB file has all b-factors in the 100s.
>
> Whazzup with that?
>
> Thanks,
> Scott