[phenixbb] strategy of refinement for a structure phased by bromide soaking

[Pascal EGEA]

Dear All,
I have a 1.8A resolution data set of a protein collected on a crystal  
soaked in KBr and I  am solving using Phenix.
I used AutoSol and gave a rough estimate of f_prime and  
f_double_prime for the Br scatterer as a starting point.
It worked great and Phenix found 31 Bromide ions (only one is fully  
occupied) and the resulting map is remarkably clear. Automatic  
building has done a great job and I was able to build most of the 4  
monomers present in the AU.

However I am confused and have a couple of questions.
1-As Phenix started it  generated a  
fobs_experimental_phases_free_Rflag.mtz file. I have been using this  
file and no longer my data.hkl file for further refinement.
I would like to know why the best_solution after the automatic  
building run does not contain any of the Bromide ions. It has water  
molecules but no Bromide ions? What is the reason for that?

2-I have been refining against the mtz file I mentioned but without  
any of these Bromide ions and although the refinement is converging ,  
I suspect I have been doing something bad and wrong.
Should I start my refinement again from the first best pdb after  
automatic building and include right away the bromide ions with  
simultaneous f_prime, f_double_prime and occupancy refinement for the  
anomalous scatterers?

As it stands now the Rfree is 22.9 and the Rfac is 19.7. The map  
looks fine except for the solvent, some of this atoms are obviously  
bromide and there is extra density in the Fo-Fc around some "solvent"  
atoms.

Thanks a lot in advance for your suggestions.

Pascal Egea
University of California San Francisco
Department of Biochemistry and Biophysics