[phenixbb] anisotropic ADP factor refinement and optimize_wxc

Pavel Afonine PAfonine at lbl.gov
Tue Jul 21 21:34:31 PDT 2009

Hi Gordon,

the target for ADP (B-factor) refinement is :

T = wxu * wxu_scale * Txray + wu * Tadp

and the target for coordinate refinement is:

T = wxc * wxc_scale * Txray + wc * Tstereochemistry .

The weights wxc (wxu) are determined automatically (similarly to CNS) as 
described in:

Adams, P.D., Pannu, N.S., Read, R.J. & Brünger, A.T. (1997). Proc. Natl. 
Acad. Sci. 94, 5018-5023;

the weight scale wxc_scale (wxu_scale) is 0.5 or 1.0 and the user can 
adjust it, wc (wu) = 1 (and the only purpose of it is to be able to turn 
off the restraints), Txray is x-ray target function (maximum-likelihood, 
least-squares, mlhl, ...), and Tadp (Tstereochemistry) is restraints term.

When you use "optimize_wxc=true" you optimize weights for coordinate 
refinement, and to do similar thing for ADP refinement you need to use 

> phenix.refine protein1.pdb 072109checks.mtz ordered_solvent=true strategy=individual_adp adp.individual.anisotropic="not element H" optimize_wxc=true --overwrite

This command above refines only anisotropic B-factors of non-H atoms, 
while the coordinates are not refined at all, and therefore the 
statement "optimize_wxc=true" is not used.
I guess what you really meant is:

phenix.refine protein1.pdb 072109checks.mtz ordered_solvent=true 
adp.individual.anisotropic="not element H" 
adp.individual.isotropic="element H" optimize_wxc=true optimize_wxu=true 

By the way, what is the resolution of the data?

If the geometry is still not good (it's another page to write to 
describe why it is potentially possible), then instead of using 
"optimize_wxc=true" try manually set up the weight putting less weight 
on X-ray term: use wxc_scale=XXX for this, where XXX is a value less 
than the default one (0.5). Say, try 0.2, 0.1, etc.

> Is it possible to refine both anisotropic B factors and optimize ADP factors in the same run.

Yes. You can refine any set of parameters for any selected part of your 
structure in any combination.

If it is still not clear, please fell free to send me the model and data 
and I will get back to you with the optimal refinement strategy that you 
can use in future as the example.


More information about the phenixbb mailing list