Chen,
First of all, it sounds funny that you can't integrate your data in C2 or P2.
That should be possible if your lattice is hP. But then if your Rmrg is OK in
P622, p622 is probably at least your (apparent) point group.
Second, I'm usually suspicious of SG P622. I would almost bet you're missing
at least one screw (that happens easily, especialy with hexagonal plates on
single axis collection; a very long z-axis, e.g. is usually a good indication
of P6(1/5)22). Check the end of the pointless output to make sure the
assignement is not done because of 'no observations'.
Try phaser in different runs with all possible spacegroups individually
selected. If you get (for some reason) high Z-scores for multiple space groups
early on, the 'automatic' space group assignement often fails, especially with
high NCS.
Don't dismiss twinning based on not beeing able to detect it. Some structures
e.g. are hypercentric and present 'normal' intensity statistics when twinned.
Some types of NCS could also mask twinning. Figuring that out requires more
work. You basically have to scale your data in all possible 'real' symmetries
(assuming merohedral twinning), do your replacement and define the 'left out'
symop as twinlaw. But I guess when your merging stats are fine you most
probably have a perfect twin (if you have one at all) and your refinement will
suck anyways.
Providing cell constants, pointless log (both for SG assignement), truncate log (maybe in comparision with your C222(1) dataset; for twinning) would help discerning between many of those possibilities.
Good luck,
Jens
-----Original message-----
From: Chen Guttman <guttm...@bgu.ac.il>
To: CCP4BB@JISCMAIL.AC.UK
Sent: 1970 Jan, Thu, 1 00:00:00 GMT+00:00
Subject: [ccp4bb] Difficult to solve MR - A mysterious case
Dear Colleagues,
Im on the bring of giving up on a very difficult and puzzling structure and
my last hope lies here with you folks...
Background:
Im working on a point mutated protein with a known w.t. structure. This
specific mutation have been crystallized in the form of plates (C2221 SG)
and hexagons (P622 SG). Whilst the plate form structure was solved without
of a hiccup, the later was a different story. I am attaching two images of
the obtained diffractions of the hexagon crystal, 90 degrees apart (the 905
images were collected at a synchrotron beamline with a delta phi of 0.2).
What I've done so far:
- Diffracted images were indexed & integrated with iMosflm (used
pointless to asses the correct SG). This was later scaled in SCALA to 2.4A
with the attached statistics.
- HKL2000 was also used but it couldn't lock on the P622 symmetry (gave
either C2 or P2 which fit to the first section of the data set but didn't
fit the 90 degree diffraction pattern); Splitting the data into two data
sets at the 90degree gave a P6 option but once it finished the first 90
degree, it missed-fitting the 2nd data set.
- Scaled data was phased using Phaser with the known w.t. structure which
gave a single solution with the following statistics:
RFZ=7.6 TFZ=33.4 PAK=0 LLG=1316 LLG=1426 and with an initial R-factor of
48.7
- Looking at density map, I could distinguish areas which had excellent
fit while other areas demonstrated poor fit. The ligand, which was soaked
with the crystal and was also solved with the w.t. protein, fitted almost
perfectly with the identified positive peaks. However, apparently there were
also major changes (mainly positive peaks) that could not be accounted for
by the structural restraints and were hard to figure out. These major
changes, which might as well be ghosts of a problematic data set, are the
reason I want to solve this structure
- Rigid body, Restrained refinement and real space refinement didn't
improve Rwork/Rfree beyond 0.4/0.46.
I've also tried the following steps:
- Used AMore and Molrep - didn't yield any improvement in density map nor
in the R factor statistics
- Used Phenix Refine with different settings
including simulated annealing, didn't work either
- Tried rebuilding using Autosol and generation of OMIT map - this
yielded some sort of improvement to 0.39/0.44 but still the map was not
drastically improved and there where areas which were hard to fit with the
current model.
- Tried removing stretches of residues and check the map after refinement
- mostly i've noticed spots of positive peaks related to the backbone.
- I've also checked for wrong use of symmetry with several programs,
including Zanuda, and still the best SG is P622.
- I've checked for twining with Phenix, CCP4 and web hosted programs and
no twining was detected (not even perfect twining).
So, I've pretty much hit a brick wall. I would be happy to hear any
suggestion to what might be the problem with this Data set.
Thanks for the help and time,
Chen
--
Chen Guttman
The Zarivach laboratory, Building 39, Room 009B
Ben-Gurion University of the Negev
POBox 653
Zip Code 84105
Beer-Sheva
Israel
http://lifeserv.bgu.ac.il/wb/zarivach
Tel. +972-8-6479519
Fax. +972-8-6472970
