*** For details on how to be removed from this list visit the ***
*** CCP4 home page http://www.ccp4.ac.uk ***
Dear Mario,
There are certainly circumstances in which the mosaicity
refinement in MOSFLM is not very stable, and normally this results in
the mosaicity refining to zero or a very small value. It is quite
unusual for it to refine to too large a value. This instability is
related in many cases to having split spots, reflecting the fact that
there are two (or more) crystals present in slightly different
orientations, and in this case the model used to refine the mosaicity
is not correct and the refinement is unstable. The best approach in
such cases is the fix the mosaic spread at a value estimated by visual
inspection of the images as you have already done (and if necessary
repeat the processing trying different values and comparing the
merging statistics from SCALA). However, there are also cases where
the mosaic spread refines to zero where there is no evidence of split
spots in the images. The reason for this is not clear at present and
is under investigation.
If the mosaic spread is small, then it can also refine to zero if
there are significant errors in the cell parameters. Lysozyme is a
good example of this. If the initial estimate of the cell (from the
autoindexing) is poor, then when the cell is refined (by
post-refinement) the mosaic spread (refined at the same time) can
refine to a value close to zero in the first cycle of refinement, but
as the cell improves then it will refine to a sensible value.
I have not personally seen an example of the mosaic spread refining to
too high a value so I cannot comment on this.
To answer your questions explicitly:
1) You can use the POSTREF MOSADD keywords to add a "safety factor" to
the refined value, but if the mosaic spread refinement is not stable
this will not necessarily give you the result you want. If you really
think the mosaic spread is changing with phi (image number), then the
only way to deal with this is to process the images in a number of
sections specifying a different mosaic spread for each, eg:
POSTREF FIX MOSAIC
MOSAIC 0.6
PROCESS 1 TO 20
GO
MOSAIC 0.7
PROCESS 21 TO 40
GO
MOSAIC 0.8
PROCESS 41 TO 60
GO
etc etc
Be sure to check visually on the images that the values for the mosaic
spread are correct.
2. By default the program will reject reflections that have a width
(in phi) of more than 5 degrees. This can be reset with the keyword
MAXWIDTH, eg:
MAXWIDTH 8
will allow reflections with widths up to 8 degrees. You will also need
to change the corresponding keyword in SCALA. There is an upper limit
of 100 images in MOSFLM, so in your case you should not try to set
MAXWIDTH to more than 10 degrees.
3. Ideally, the Rmerge should be lower for fine phi sliced data, but
this is only true up to a point because each image has its own
detector noise and possibly systematic errors due to shutter
synchronisation. As others have said already, there is generally
little to be gained by using slices less than one third to one fifth
of the mosaic spread (plus beam divergence). Of course, the Rmerge
will also depend on the strength of diffraction, so I'm not sure what
exactly you mean by "higher than usual".
4. I find the correlation coefficient between the anomalous signal for
two random halves of the dataset much more informative than the normal
plot. I think this is in the latest version of SCALA (it has been in
our lab version for some time).
The table looks like this:
$TABLE: Correlations within dataset, Hg :
$GRAPHS: Anom & Imean CCs v resolution - :A:2,4,6,12:
:RMS correlation ratio:A:2,8,10:$$
N 1/resol^2 dmax CC_anom N_anom CC_cen N_cen RCR_anom N_anom RCR_cen
N_cen CC_Imean N_Imean $$
$$
1 0.0199 7.08 0.864 78 0.000 0 4.513 77 0.000
0 0.999 78
2 0.0399 5.01 0.839 120 0.000 0 3.363 120 0.000
0 0.997 120
3 0.0598 4.09 0.760 144 0.000 0 2.709 144 0.000
0 0.997 144
4 0.0797 3.54 0.644 138 0.000 0 2.133 138 0.000
0 0.997 138
5 0.0997 3.17 0.692 145 0.000 0 2.939 142 0.000
0 0.999 145
6 0.1196 2.89 0.758 129 0.000 0 2.660 129 0.000
0 0.997 129
7 0.1395 2.68 0.663 125 0.000 0 2.206 125 0.000
0 0.993 125
8 0.1595 2.50 0.525 116 0.000 0 1.799 116 0.000
0 0.991 116
9 0.1794 2.36 0.534 102 0.000 0 1.678 101 0.000
0 0.991 102
10 0.1993 2.24 0.577 60 0.000 0 1.799 58 0.000
0 0.865 60
$$
Overall 0.748 1157 0.000 0 2.790 1150 0.000
0 0.998 1157
This is for a stunningly good anomalous signal !
Best wishes,
Andrew Leslie
