[phenixbb] some basic questions about phenix autobuild and phenix.refine

Nathaniel Echols nechols at lbl.gov
Wed Oct 24 08:24:22 PDT 2012

On Wed, Oct 24, 2012 at 12:37 AM, Niks <nikdna at gmail.com> wrote:
> 1) I have solved my structure using AutoMR and got two files MR.1.mtz and
> MR.1.pdb as output with high LLG of 8236.889. When going for AutoBuild
> afterwards it uses output pdb (MR.1.pdb) which is fine but uses MR.1.mtz as
> map coefficient. Shouldnot it uses original mtz from scala?

It should use the original MTZ file for the experimental data - but
since Phaser already calculates a weighted map, it is most convenient
to use that to start the building process.  Which version of Phenix
are you using?  Can you send me a screenshot of the input window?

> 2) Same in Phenix.refine when going after Autobuild finishes and it asks for
> refining the model it uses same AutoMR output mtz (MR.1.mtz). As I was
> suggested by many that we should use original mtz coming from scala for
> Refinement to reduce model bias. shouldnot phenix.refine use Original mtz
> file here too?

Unless AutoBuild used MR.1.mtz as the data file too, that sounds like
it could be a bug.  I will check it out.

> 3) After going for AutoBuild after AutoMR, even when I unselect "Place
> Waters" options in parameters for Refinement, I see output pdb file with
> waters added to it. I do not want waters to be added till I refine my model
> to the best , but Autobuild option comes out with waters added. Is there
> anyway I can have autobuilt pdb without waters added?

Did you remove the waters from the input file before running MR?  If
so, this also sounds like a bug.

> 4) Another question that troubling me most that  in Phenix.refine even if I
> use original mtz which clearly specifies resoultion range from 127.488-2.5A
> my refined pdb files comes out with  Resolution range used for refinement as
> 37.8-2.5A. I have tried to input low resolution range as 127.488 but deafult
> is always 37.8 selected. My data may be very poor in lower resolutions than
> 37.8A but my completeness is also going down (65.5% as opposed  to 79.8% in
> REFMAC). Can I select my refinement using my specified values of Resolution
> from mtz file coming out from SCALA?

I think the discrepancy is due to Phenix reporting the resolution for
reflections which are actually measured.  The MTZ file might claim
that it goes from 127.5 Å down, but it probably has HKL indices (and
maybe R-free flags) for those reflections between 127-38Å but no
actual amplitudes or intensities.  Once the data are extracted into
Phenix it discards the reflections for which no data are available,
and reports only the resolution it sees.

The discrepancy in completeness statistic could happen if you're using
anomalous data in Phenix but merged data in Refmac; Phenix will report
the completeness keeping F+ and F- separate.  I'm more worried about
the low completeness in general, however; unless you just have very
incomplete high-resolution shells, this much missing data tends to
result in map artifacts and other difficulties during refinement,
especially if it's a systematic chunk of reciprocal space that's
missing (as is usually the case).  The "3D data viewer" under
"Reflection tools" will let you visualize which data are missing - it
might be worth checking this out.


More information about the phenixbb mailing list