Re: [phenixbb] Refining f'/f"?

7 May 2009

      Via the U. Washington anomalous scattering tables for Selenium, since 
1.0809 = 11471 eV
f' ~ -2.
http://skuld.bmsc.washington.edu/scatter/data/Se.dat
Since Se has 34 electrons one could knock the occupancy down to 0.94 and 
see if that helps the maps.  Radiation damage might not be as focussed 
on the Se away from the edge, but one could experiment with other 
occupancy values and see if that brings the B-factor of the Se in line 
with the CG atom - there's the f' correction, radiation damage and the 
partial SeMet incorporation all playing a role here.  Of course, Mets 
have a fair degree of conformational flexibility so this is an imperfect 
solution since partial occupancy models a whole range of sins.

I don't have a handle on the relative contribution to net f" of Se and 
the far more numerous C/N/O/S at this wavelength, but I think you'd have 
to consider modeling those as anomalous scatterers too if treating F(+) 
separate from F(-).

One note concerning Scalepack, from reading Nate's post (which overlaps 
mine).  The "anomalous" flag treats I(+) = I(-) from the scaling point 
of view but outputs them separately.  "Scale anomalous" treats I(+) and 
I(-) independently during scaling and as I suspect you're in a high 
symmetry point group (422) with reasonable redundancy I'd suggest using 
the latter.

Phil Jeffrey
Princeton

Leigh Allen wrote:
...
I collected at 1.0809A, which was the optimal wavelength for the beam at 
NSLS-X29.  I will model them as MSEs.  I guess the better question to 
ask is if I should be concerned with not having these negative peaks.  
Will it negatively affect my R-work/R-free if I continue to work with 
the data that's not been scaled to keep F+/F- separate?
-Leigh
On Thu, May 7, 2009 at 3:10 PM, Nathaniel Echols mailto:[email protected]> wrote:
On May 7, 2009, at 2:53 PM, Leigh Allen wrote:
     > I'm new to the world of x-ray crystallography.  I just solved my
     > first SAD structure and I'm on to the refinement stage.  The
     > difference map generated by phenix.refine has really big positive
     > peaks all around my MET residues.  If I switch the atoms to MSE,
     > then I get large negative peaks.  Firstly, I'm not sure if I'm
     > supposed to represent these residues as MET or MSE because it's a
     > "native," high resolution (1.85) dataset, but it the protein had MSE
     > residues.  When scaling this data, I did not keep F(+) and F(-)
     > separate.  The protein's phases were generated using SAD data to
     > 2.7A, which using SHARP led to a remarkably interpretable map that
     > allowed me to build in the protein by hand. My second question is,
     > how should I handle the issue with large + MET/large - MSE peaks?
     > Do I need to rescale my data to treat it as anomalous data or is
     > there something I can do within Phenix to fix my problem.  I tried
     > the phenix.refine GUI and set it up to refine f' and f", but it
     > appears that nothing really changed.
When you wrote "native", do you mean collected at a wavelength
    significantly different than the Se K edge?  If it's a longer
    wavelength, there will be much less anomalous signal anyway.  However,
    without separate Friedel pairs I think it is impossible to tell, so if
    you want to refine the anomalous coefficients you should rescale with F
    +/F- kept separate.
Regardless of the anomalous signal, if the protein contained Se you
    should model the METs as MSEs.  I think it's very common for these to
    have negative difference map peaks, because they'll undergo radiation
    damage very quickly relative to the rest of the protein.  If you
    collected the high-resolution data set solely to get high resolution
    and not phases, this is probably what happened, but I've even seen the
    negative peaks around sites used for phasing in a low-exposure dataset.
-Nat
    _______________________________________________
    phenixbb mailing list
    [email protected] mailto:[email protected]
    http://www.phenix-online.org/mailman/listinfo/phenixbb
------------------------------------------------------------------------
_______________________________________________
phenixbb mailing list
[email protected]
http://www.phenix-online.org/mailman/listinfo/phenixbb