From: Ian Tickle
Date: 27 October 2011 20:31
Hi, currently Refmac writes the wrong info for the low resolution
cutoff, for example:
REMARK 3 PROGRAM : REFMAC 5.6.0119
REMARK 3 DATA USED IN REFINEMENT.
REMARK 3 RESOLUTION RANGE HIGH (ANGSTROMS) : 1.00
REMARK 3 RESOLUTION RANGE LOW (ANGSTROMS) : 84.35
In contrast Shel-X puts in:
REMARK 3 RESOLUTION RANGE LOW (ANGSTROMS) : 10.00
The mtzdump of the MTZ file shows:
1 NONE -83 0 0 100.00 -27.0 27.0 84.35 1.00 H H
2 NONE 0 97 0 100.00 54.9 54.9 84.35 1.00 H K
3 ASC 0 83 0 100.00 31.0 31.0 84.35 1.00 H L
4 NONE 0.0 1.0 0 100.00 0.95 0.95 84.35 1.00 I FREE
5 NONE 0.0 2071.8 435 99.82 125.22 125.22 10.00 1.00 F FP
6 NONE 0.9 90.0 435 99.82 2.73 2.73 10.00 1.00
Q SIGFP
So obviously what has happened is that the Free R flags have been
generated to the high resolution cutoff and the low-res reflections,
i.e. below 10 Ang only have HKL & FREE defined. Somewhere along the
line (this is data from the PDB) the low-res F's were lost. So IMO
one cannot claim that these data were "used" in refinement. Getting
tthe low-res cutoff wrong has a big effect on the electron density
that one expects for a given resolution range! It seems to me that it
would be preferable to omit information rather than risk putting in
the wrong value, at least then one would know to go find the info
somewhere else.
Cheers
-- Ian
----------
From: Tim Gruene
Dear Ian,
"10.00A" sounds very much like somebody used shelxpro to create an
ins-file from a PDB file and did not correct the SHEL card to include
low resolution data. So the data may well contain data at 84A, but shelx
does not use the data if the SHEL card is left at its default (10 0.1).
Cheers,
Tim
P.S.: "obvious" can be just a dangerous word as "trivial"...
- --
- --
Dr Tim Gruene
Institut fuer anorganische Chemie
Tammannstr. 4
D-37077 Goettingen
----------
From: George M. Sheldrick
Dear Ian,
Truncating the data to 10A and then filling in the missing hkl values and
including them with Fo replaced by e.g. DFc is an established way of
improving the maps (at the cost of a little model bias), but the novel
twist in your example was the assignment of R-free flags to these missing
reflections. Does this also lead to a reduction in the free R?!
Best wishes, George
--
Prof. George M. Sheldrick FRS
Dept. Structural Chemistry,
University of Goettingen,
Tammannstr. 4,
D37077 Goettingen, Germany
No comments:
Post a Comment