Hi Yuan Cheng,
You have received plenty of advice as to how to proceed with your
problem. There is still one thing that no one mentioned, and in my
opinion, is the simplest way to go about what you want to do.
There are plenty of monoclinic cells with a beta that is very close
to 90, but when you start with orthorhombic, you don't necessarily
know which axis is the unique one. So I would go back to the
beginning, and at the autoindexing step, choose the same cell
setting for both crystals, whichever space group you choose. You
have already identified that a and b had been switched. You don't
have to live with that. You can choose a different setting when
autoindexing gives you the options.
Now, if you are already convinced that it is monoclinic, you still need
to know which is the correct b axis, but at least the two data sets
should be internally consistent, i.e. R-merge ought to be similar for
the overall data set as it is for the individual one, usually an average
of the two. Deciding on the correct b axis will have to be done in a
different way. Usually MR (I assume you do have a starting model)
is very good at selecting the correct setting. You start with your
chosen setting, and if that gives you a good solution, then you have
got the correct setting very likely. Just to convince yourself, you can
try the other setting. It could much better or much worse. It rarely
is about the same, because that would require a tetramer in the
asymmetric unit, which has a 4-fold local symmetry. I doubt this
is what you are dealing with. If I understand the gist of your message
you have 2 copies in the a.u.
Good Luck.
Pierre
**********************************************************************
Dr. Pierre Rizkallah, Senior Lecturer in Structural Biology, WHRI, School of
Medicine, Academic Avenue, Heath Park, Cardiff CF14 4XN
email: rizkall...@cf.ac.uk phone + 44 29 2074 2248
>>> Yuan Cheng <ych...@email.unc.edu> 22/10/09 9:52 PM >>>
Hi everyone,
I have some question about the merging of diffraction data from
different crystals. I have two (room-temperature) datasets in my hand
and each of them can be scaled in P21 space group and the screw axis is
along k. However,each dataset only have below 55% completeness. To get
better completeness, I am trying to merge this two datasets. The problem
is one dataset has cell dimension (80.632,87.085,114.977),(90,90.026,90)
and the other one has (85.497,79.857,114.003) (90,90.004,90),. It looks
like that the a,b dimensions are switched between these two datasets.
Unsurprisingly,the chi^2 is very high when I tried to merge these two
datasets in scalepack.I am wondering whether there is any way to reindex
these two datasets to make their a/b dimensions match. Any suggestion
will be highly appreciated.
By the way, as you might notice, the beta angle is pretty close to
90 degree. I scaled the datasets into P212121 or P2221 or P21212 at the
beginning,I had no trouble to merge them in these space groups.
However,I could not make the Rfree go below 45% during following
refinement (2.7 angstrom cutoff). Phenix.xtriage indicated that there
might exist twinning but no twin law is given. Then I reindexed the data
into P2 and scaled them in P21 SG.Phenix.xtriage indicates there exists
a pseudomerohedral twinning operator. When I used the twin law given in
phenix.refine,the Rfree could go down to 29%. So I think P21 might be
the correct space group.
Thanks again for any suggestion.
Yuan
