Whilst we are on the subject of XDS...
I had difficulty processing a data-set in mosflm the other day so on
the recommendation of a colleague switched to xds which, with a bit of
tweaking, seemed to work really well. I converted the resulting
XDS_ASCII.HKL using xdsconv and then f2mtz ready for phaser and refmac.
However, my colleague then told me that xds handled missing
reflections differently from the usual mosflm/CCP4 route and thus I
had to run the CCP4 program UNIQUE before I tried refinement as
apparently refmac does not like reflection files originally processed
with xds. As I couldn't find anything in the literature about this I
was wondering whether this advice is still up to date?
Thanks,
Simon
On 21 Feb 2008, at 09:44, Kay Diederichs wrote:
Engin Ozkan schrieb:
Hi everyone,
I have been recently relying on XDS quite a bit, but at the same
time worrying about how XDS treats overlaps. We had one dataset
that both HKL2000 and Mosflm would show to have severe overlaps, as
expected due to unit cell parameters and the unfortunate crystal
orientation in the loop. We always ended up with completeness
percentages in the 70's.
XDS can find the same lattice, index and scale the data, but yields
a 100% complete mtz (and a nice structure). Without the HKL/Mosflm-
like GUI, it is difficult to assess the fate of the overlapped
observations in XDS. What I could see with VIEW was that some
observations were being divided into several ovals, probably
different reflections, but I'm not very certain.
So, the basic question is, how does XDS treat overlaps? I could
not find in the documentation an answer to this question; the
single mention of overlaps I could find tells me that XDS can
recognize overlaps, but does not tell me if it rejects them, or
divvies them up into separate reflections, and if that is the case,
how does it divide them, and how reliable is that? Depending on how
it divides the overlaps, could that affect commonly-used intensity
stats and distributions?
Thanks,
Engin
Engin,
the basic answer is:
a) each pixel of the detector is assigned to its nearest reflection
in reciprocal space
b) some of these pixels will mostly allow the background estimation,
others will mostly contribute to the integration area (but as they
are transformed into a local coordinate system there is not a 1:1
relationship). At this step, pixels which should be background but
are higher than expected (due to overlap) are rejected.
c) for each reflection, the background is estimated, and the 3D
profile is assembled from the pixels contributing to it
d) a comparison is made: for a reflection, is the percentage of its
observed profile assembled in c) larger than some constant (called
"MINPK" in XDS.INP)? If the answer is no, this reflection will be
discarded (you could call this situation "overlap").
Among other things, this means that:
a) the program does _not_ look around each reflection to detect an
overlap situation, it just tries to gather the pixels for each
reflection
b) as a user, when your crystal-detector distance was chosen too low
or the reflections are very broad (resulting in generally strong
overlap), you may reduce MINPK down to 50. This will result in more
completeness, but you should monitor the quality of the resulting
data. Conversely, if you raise MINPK over its default of 75 you will
discard more reflections, but the resulting dataset will be a bit
cleaner.
The reference is
W. Kabsch (1988) Evaluation of single-crystal X-ray diffraction
data from a position-sensitive detector. J. Appl. Cryst. 21,
916-924. (http://dx.doi.org/10.1107/S0021889888007903)
HTH,
Kay
--
Kay Diederichs http://strucbio.biologie.uni-konstanz.de
email: [EMAIL PROTECTED] Tel +49 7531 88 4049 Fax 3183
Fachbereich Biologie, Universität Konstanz, Box M647, D-78457 Konstanz
