Dear George,
you are correct - this kind of refinement only makes sense if F+ and
F- are kept unmerged. The refinement of f'/f'' is performed with the FFT
(from what I recall - Ralf can confirm or deny). At the very least,
structure factors from non-anomalous atoms are calculated using the FFT,
and the anomalous atoms by direct summation. Either way, the inclusion
of anomalous atoms doesn't add any perceptible cost to the refinement.
The multi-wavelength f'/f'' refinement in SHELX sounds very
interesting. We currently cannot do that in phenix.refine (refinement is
limited to one dataset, thus one wavelength). Although it may be
implemented in the future.
Cheers,
Paul
George M. Sheldrick wrote:
Dear Paul,
That implies that phenix refine (like SHELXL, but unlike REFMAC) can
refine against data for which I+ and I- (or F+ and F-) have not been
merged. An attempt to refine f" against Friedel merged data would be
very poorly determined. Am I right in guessing that phenix refine can
only do this when then structure factors are calculated by the
classical summation and not by FFT?
Given relatively high resolution MAD data, it is at least theoretically
possible using SHELXL (and presumably also phenix refine) to refine
against I+ and I- for all wavelengths at once, this should give much
better occupancies for the selenium atoms (in the case of SHELXL it
would be necessary to use HKLF 2 input format so that a wavelength is
specified for each reflection, refine batch scale factors, use the
LAUE instruction so that the program can work out f' and f" for each
reflection, and - at least when each site is assumed to be Se or S in
different ratios - use 'free variables' for the Se occupancies). If anyone
is MAD enough to try this, I would also recommend scaling against a
reference dataset (e.g. high energy remote) rather than scaling the data
for different wavelengths separately. There is a suitable option in SCALA
and SADABS and possibly in other scaling programs as well.
Best wishes, George
Prof. George M. Sheldrick FRS
Dept. Structural Chemistry,
University of Goettingen,
Tammannstr. 4,
D37077 Goettingen, Germany
Tel. +49-551-39-3021 or -3068
Fax. +49-551-39-2582
On Sat, 11 Aug 2007, Paul Adams wrote:
Hi Ethan,
if you want you can specify and refine both f' and f'' for any user selected
atoms in phenix.refine:
http://www.phenix-online.org/download/
Cheers,
Paul
On Aug 11, 2007, at 9:48 PM, Ethan A Merritt wrote:
On Saturday 11 August 2007 21:41, Miller, Mitchell D. wrote:
Hi Ethan,
I have been wanting a way to instruct refmac to accept a user-
defined
f' term since about forever.
According to Garib's latest release notes, a command was
added to refmac 5.3.0015 and later to allow f' to be specified.
I have not tried it myself yet.. see
http://www.ysbl.york.ac.uk/~garib/refmac/data/refmac_news.html
"Anomolous form factors
anom formfactor [Name] [f'] [f'']
It will modify form factor of the given atom and use f' part only"
<goes off to install a newer refmac version/>
Thanks,
Ethan
--
Paul Adams
Senior Scientist, Physical Biosciences Division, Lawrence Berkeley Laboratory
Adjunct Professor, Department of Bioengineering, U.C. Berkeley
Head, Berkeley Center for Structural Biology
Deputy Principal Investigator, Berkeley Structural Genomics Center
Building 64, Room 248
Tel: 510-486-4225, Fax: 510-486-5909
http://cci.lbl.gov/paul
Lawrence Berkeley Laboratory
1 Cyclotron Road
BLDG 64R0121
Berkeley, CA 94720, USA.
--
--
Paul Adams
Senior Scientist, Physical Biosciences Division, Lawrence Berkeley Lab
Adjunct Professor, Department of Bioengineering, U.C. Berkeley
Head, Berkeley Center for Structural Biology
Deputy Principal Investigator, Berkeley Structural Genomics Center
Building 64, Room 248
Tel: 510-486-4225, Fax: 510-486-5909
http://cci.lbl.gov/
Lawrence Berkeley Laboratory
1 Cyclotron Road
BLDG 64R0121
Berkeley, CA 94720, USA.
--
