[phenixbb] question about refinement in a SAD data set

Thomas C. Terwilliger terwilliger at lanl.gov
Sun Mar 30 16:24:44 PDT 2008

Hi Pascal,

> Dear All,
> I am trying to refine a seleno met structures at 2.2 Ang resolution
> I used Auto_Sol and Auto_Build and it worked great since it found the
> 4 semets with occupancies of about 0.85 to 0.95. and most of the
> model was built. Phenix generated  an exptl_fp_phases.mtz with the
> Rfree flag. so far so good. I have a couple of questions because I am
> not quite to understand what phenix.refine is doing at this stage?
> 1- If I refine  using the experimental phases mtz file generated by
> the auto_sol run, the occupancies for the seleniums change quite a
> bit and they drop to 0.75-0.80 and that does not sound right to me?
> What am I doing wrong? should I fix occupancies for the seleniums to
> the values evaluated at the end of the auto_build run (Rfree=27/
> Rfac=24?)

The file exptl_fobs_phases_freeR_flags.mtz from AutoBuild will normally be
(essentially or actually) identical to the file
exptl_fobs_phases_freeR_flags_2.mtz from AutoSol, where  2 refers to
solution 2. Whatever the best solution was from AutoSol, that
exptl_fobs_phases_freeR_flags_2.mtz gets copied to
exptl_fobs_phases_freeR_flags.mtz for AutoBuild.

I would therefore have expected that the refinement at the end of
AutoBuild, which refined your best PDB file against
exptl_fobs_phases_freeR_flags.mtz to yield overall_best.pdb, would give
essentially the same result as refining overall_best.pdb against the
AutoSol exptl_fobs_phases_freeR_flags_2.mtz file.  So yes, I am a little
surprised too.

However occupancies of 0.75-0.80 for Se atoms is not unusual. This
occupancy can come from a mixture of S and Se at these locations.  I would
let the Se occupancies refine.

> 2- This experimental phases mtz file contains FP and the HL
> coefficients for the phases but there are no more anomalous
> differences.   If I try to refine f' and f" using it,  Phenix tells
> me that there is no anomalous information contained in this file. So
> how do you refine f' and f"? and what is the practical reason for
> this AutoSol generated mtz file to not contain the anomalous
> differences that were used to locate the seleniums.

Yes, there are no anomalous differences in the
exptl_fobs_phases_freeR_flags.mtz files.  This will change soon when
phenix.refine gets the ability to refine using a SAD function. At the
moment including the anomalous differences in refinement isn't likely to
change much at all in your structure.

If you want to use anomalous differences in refinement right now, then you
can run phenix.refine with your FreeR_flags coming from
exptl_fobs_phases_freeR_flags.mtz and your I+ and I- coming from your
starting data file (sad.sca or equivalent).

All the best,
Tom T

More information about the phenixbb mailing list