Hello All,
I have a question about some data with which I have been grappling. The crystal
appears to be primitive cubic(p23). Are there alternate ways to index cubic
data? I ask this question without regard to anomalous data. I have looked at
several datasets in Xtriage. I wanted to merge data from two crystals together
and I looked to see if it were necessary to reindex to do it properly. Here is
what xtriage has to say.
Merging analysis between datasets from 2 different crystals(These datasets are
"non-anomalous" datasets, ie Bijvoet pairs have been merged).
Reference analyses
The following reindexing operators have been found:
-------------------------------------------------
| Operator | Correlation | matches (%) | choice |
-------------------------------------------------
| h,k,l | 0.000 | 89.6 | |
| -h,l,k | 0.722 | 91.4 | <--- |
-------------------------------------------------
I'd have thought that there was a single indexing possibilty for the cubic
space group in the absence of a consideration of the anomalous signal. And I'd
accept that I were wrong if I hadn't looked at some other data.
So to cloud my thinking, I decided to look at the output from Xtriage for
datasets from the same crystal collected 180 degrees apart. In this case the
data were processed with HKL2000 with the same indexing setup, ie the crystal
rotx,y,z reference zone etc and starting phis of x and x+180 were used. The
data in this case is again treated as "non-anomalous" too.
Reference analyses
The following reindexing operators have been found:
-------------------------------------------------
| Operator | Correlation | matches (%) | choice |
-------------------------------------------------
| h,k,l | -0.007 | 91.0 | |
| -h,l,k | 0.719 | 92.7 | <--- |
-------------------------------------------------
What's going on here? I processed these two wedges at different times, but I
kept all of the parameters consistent. Are there two choices of indexing? this
doesn't seem possible, given that the indexing set-up was the same. or does
denzo arbitrarily choose the configuration upon initial indexing irrelevant of
what I tell it? I think not. Or might this show that my space group is of lower
symmetry(rhombohedral or orthorhombic) than p23, with or without some possible
pseudo-symmetry?
In addition, I have other crystals of a similar complex(one protein has a
single aa mutation) which are I23 with similar(2% difference) in cell
constants. Same set-up as case 2, datasets from the same crystal collected 180
degrees apart, data were processed with HKL2000 with the same indexing setup,
ie the crystal rotx,y,z reference zone etc and starting phis of x and x+180
were used. The data in this case is again treated as "non-anomalous" too. This
data was processed at the same time.
Reference analyses
The following reindexing operators have been found:
-------------------------------------------------
| Operator | Correlation | matches (%) | choice |
-------------------------------------------------
| h,k,l | 0.945 | 59.7 | <--- |
| -h,l,k | 0.110 | 58.4 | |
-------------------------------------------------
This is what I'd have expected for case 2 as well.
Lastly, for the P23(nor the I23 case for that matter) case, Xtriage doesn't
indicate twinning. In each case, I can get a molecular replacement solution
which has very reasonable near identical packing, my search model is 100%
identical to one of the 2 proteins in the crystal. Unfortunately I don't see
the second protein in the maps. It is about 20 percent the size of the other
protein. The obvious answer is the second protein isn't there but I feel that
it should be there because I get a nice Se fluorescence signal from the
crystals, and the missing protein is the only one that is labeled. So I'm
looking at possibilities of a misassigned space group, etc to get over the hump.
I hope all of this doesn't sound like mindless drivel. Any comments on any or
all of this?
Thanks in advance.
todd
