[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ccp4bb]: Twinning/Merging



***  For details on how to be removed from this list visit the  ***
***    CCP4 home page http://www.dl.ac.uk/CCP/CCP4/main.html    ***

Daniel,

I guess with "Rmerge" you mean what is also sometimes called "Rsym", but
applied to a dataset merged from several crystals? First of all, Rmerge
is a poor indicator of dataset quality. There are better ones available
(see Nature Structural Biology 4, 269 (1997)), but regrettably SCALEPACK
does not calculate them (CCP4 SCALA and XDS/XSCALE do). Using Rmerge for
basing decisions about e.g. inclusion of datasets into a merged dataset
can be quite misleading!

There are two possible reasons for high Rmerge:
1) bad counting statistics resulting from too short exposure, or small
crystals, or weak beam. The problem can in principle be overcome by
merging several datasets, exposing the small crystals longer, or
possibly by growing larger crystals, or going to the synchrotron, or ...
Normally, the data have a good Rmerge (5% or so) at low resolution, and
Rmerge goes up with resolution.

2) systematic differences between crystals resulting from aniosomorphism
(or - in case of merohedral twinning - different twinning fraction), or
systematic differences within each individual dataset resulting from
radiation decay (this also results in anisomorphism). Often, data with
this problem have a higher-than-usual Rmerge at low resolution (but
rejections tend to mask this), and a high number of rejections.

I do not currently understand why you suspect that twinning by itself
should increase Rmerge? Twinning changes the intensities, but it does
not break the symmetry between reflections.

16% Rmerge is not too high by itself. If it results from reason 1), the
data can be quite valuable. I don't think you will be able to get lot of
structural information from the 2.38-2.35 A shell, though. But in MLF
refinement (CNS) it shouldn't hurt to include the shell either.

You need to find out about the individual twinning fractions of the 3
crystals that made up your merged dataset. If the 16% value results from
reason 2), it would be highly desirable to know about these systematic
effects.

Your 100K data: 5-15% twinning is not too high, you can either detwin,
or use the twinning fraction within CNS or SHELXL.

Kay
-- 
Kay Diederichs                    
email: Kay.Diederichs@uni-konstanz.de   Tel: +49 7531 88 4049  Fax: 3183
Fakultät für Biologie, M656, Universität Konstanz, D-78457 Konstanz