On Nov 22, 2007, at 14:43, Kay Diederichs wrote:
Pierre Barraud schrieb:
Dear all,
I am working on a data set which is severely anisotropic, with
diffraction limits of 2.6 A along the a* and b* direction but only
3.3 A along the c* direction. I attached a screen shot of the
Anisotropic analysis (FALLOFF) output graph from SCALA.
My question is : what is the best strategy (or what are the
different strategies) in the structure resolution process (from
image integration to refinement) to make use of the maximum high
resolution information ? Actually, I am a little disorientated by
the informations I have collected.
Best regards,
Pierre
Pierre,
as you found, data integration is not harmed by anisotropy.
Well, although Kay is formally right, I would not agree completely.
Think of a 1-2 micron thin plate, very well ordered, and an
excellent almost matching micro-beam:
shooting 'edge-on' you get a diffracting volume 'along' the plate and
data to - say - 2.5 A.
shooting 'face-on' you get a much lesser diffracting volume 'through'
the plate and data to - say - 3.2 A.
(if you shoot too much to get 2.5 you kill the crystal in no-time)
In that case I prefer to process my data with successively changing
resolution limits. This way I get
some 2.5 A data, 30-40% complete, with acceptable processing
statistics (acceptable excluding completeness obviously),
and at 3.2 A 100% completeness with also good statistics. If I would
say apply the 'usual' cutoffs during scaling in a set globally
integrated
to 2.8 A, I would get a bad dataset at 2.8, with some good and some
bad hkl's and then I would be missing some lovely reflections
between 2.5-2.8.
To do this I personally like the "RESOLUTION CUTOFF n" in MOSFLM
which will only integrate where the I/sigI for that image is n (I
like 2).
A good question is how to call the above dataset: 2.5 or 3.2 .. I
would not really care, since the "2.5" map looks better.
I can see that coming as an argument with referees though, so
consider that as a pre-emptive strike ;-)
... or we better take more crystals with us and 'scan' the plate
surface to get 100% complete at 2.5 and merge .. will try.
For anisotropy also look at:
http://www.doe-mbi.ucla.edu/~sawaya/anisoscale/
and read the 2006 paper from D. Eisenberg's lab in 2006.
A.
The relevant programs for structure solution and refinement (e.g.
sharp, molrep, phaser, refmac etc.) deal explicitly with
anisotropy; I don't think there there is a compelling need to do
something fancy. Others will correct me.
If you _do_ want an anisotropic resolution cutoff, you could use
sftools. Load all your data (i.e. up to 2.6A) into sftools. Then
multiply the length of the c axis by a factor of 2.6/3.3 , and let
sftools delete all those reflections that are beyond 2.6A now. Then
restore the original c axis and write out the remaining data.
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
