Re: [ccp4bb]: refine using merged MAD data of all 3 wavelength?

[Kay Diederichs <Kay.Diederichs@uni-konstanz.de>]

***  For details on how to be removed from this list visit the  ***
***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***

Yong Xiong wrote:
> 
> ***  For details on how to be removed from this list visit the  ***
> ***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***
> 
> As title, I would like to know whether any one tried to use all 3 wavelength data merged/scaled
> together for the refinement. I tried this on one structure and it appeared to improve the
> refinement. I am not sure the theoretical/practical soundness of doing this. If the
> crystal is stable, it should give a lot more redudency to the data and help the high resolution
> bins. The anomalous/dispersion signals would be messed up but mostly they are not used in
> refinement any way. We could even give an average f' and f" for the merged data, or could we? 
> Any comments?

If you merge MAD data from different wavelengths, you will average the
isomorphous (delta-f') and anomalous (f") contributions from your
anomalous scatterers (Se/Hg/... whatever). The contributions from all
other scatterers will be averaged and thus improved by redundancy.

a) most importantly: to get an idea what the improvement in terms of
data quality is, 
   compare R_mrgdF from each wavelength with R_mrgdF from the
merged-wavelength, high-
   redundancy dataset.  (Calculate R_mrgdF with novel_r.f - link is on
my homepage - paper 
   is Diederichs & Karplus (1997) Nature Structural Biology 4, 269 ).
Maybe you could post
   the results to ccp4bb.
b) you won't be able to use the anomalous signal of your merged dataset
for anything, but 
   that would not hurt refinement unless the program can use the
information (SHELXL, 1.4A data) 
c) the delta-F' contributions from the anomalous scatterers should
produce slightly different 
   occupancies for these atoms when refining against the individual
wavelengths (or different
   B-factors if occupancies are 1.0). When the intensities from the
wavelengths are merged, the
   averaging would not be done correctly in a strict sense as averaging
would be applied to the
   intensities, not to the complex structure factors. I would not expect
this difference from 
   correctness to be noticeable, however. The occupancies (or B-factors)
after merged-wavelength
   refinement should be just about the average of the single-wl occs (or
Bs).
d) merging probably only makes sense if there is no significant
radiation damage

In summary, I'd say it would be a good idea and could be applied
generally. The test would be just easy - refine (using the exact same
protocol and starting structure) against the wls individually and
against the merged-wl dataset, and compare the R_free values. Not too
much work, and quite interesting.

Just my 2 Pfennige,

Kay
-- 
Kay Diederichs         http://strucbio.biologie.uni-konstanz.de/~kay 
email: Kay.Diederichs@uni-konstanz.de  Tel +49 7531 88 4049 Fax 3183
Fakultaet fuer Biologie, Universitaet Konstanz 
Box M656, D-78457 Konstanz, Germany