Re: [ccp4bb]: Omit maps in Xtalview

["Harry M. Greenblatt" <harry.greenblatt@weizmann.ac.il>]

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

Paul Hubbard wrote:
BS"D


> To make the map I initally calculate phases in CCP4i using current refined
> coordinates, switching the "scale input Fp to the Fcalc" toggle on. I
> make sure the working set used to make the model in CNS/O is flagged the
> same in the mtz file. Next I run the following script to make the phs
> file:
> 
> mtz2various HKLIN phased.mtz HKLOUT out.phs << eof OUTPUT USER -
>     '(3I4,x,F7.2,x,F7.2,x,F7.2)'
> labin  FP=FP FC=FC PHIC=PHIC
> end
> 
> In Xtalview I make a mFo-Fc map with cubic splines, applying bulk solvent
> correction. Finally, to make the omit map, I select the waters to be
> ignored and recalulate the map with the shake option turned on.

  There are many subtle problems here, which is why I always use CNS to
make maps, and output coefficient files for use in Xtalview.

1)  When you make a mlf/SIGMAA map in CNS, the program uses the test set
for cross validation to generate the SIGMAA coefficients.  The map
calculation uses *all* the reflections.  When you import the reflections
into Xtalview, you no longer have the test set seperate, and so the
SIGMAA calculation in Xtalview will potentially give rather different
results.

2)  In CNS, I imagine the bulk solvent correction also excludes the test
set.  Xtalview will include the test set, since it has no facilities for
FreeR.  In addition, who says they use the same method?  (Even if they
use the same method, the former point still stands.)  Furthermore, do
they use the same resolution range to calculate the correction?


> 
> Does the shake routine work as well as true S.A., and can I get
> reliable omit map; or am I missing something which is causing this density
> shift problem.

3)  Without any data on the matter, I would guess that the shake method
will not work anywhere near as well.  In SA, you heat the protein up in
a dynamics run, and then you cool it down with the xray term and
geometry terms guiding the atoms into a *refined* structure.  So you
generate a map on a refined structure, but one that might be different
from the starting model.  As far as I know, the shake method adds a
small "random" translation to each atom before the fft.  So you make are
making a map with a structure that could be termed "unrefined", since
the shifts were not applied with any consideration of the xray term. 
The R-factor of such a structure shoule be higher than the R-factor of
the starting model.  So the map will be noisier, etc.

  So stick to CNS maps...

Hope that was clear,

Harry

-- 
------------------------------------------------------------------------
  Harry M. Greenblatt           | Phone: 972-8-934-3625
  Research Assistant            | FAX  :           4159
  Dept. of Structural Biology   |
  Weizmann Institute of Science |
  Rehovot, Israel 76100         | Email: harry.greenblatt@weizmann.ac.il
------------------------------------------------------------------------