[phenixbb] Twinnig

[Michael Gorman]

Dear all,
After over 15 years solving structures I have (knowingly) encountered my 
first twinned data set.
Short history..

Processed synchrotron data with HKL2000
SG P 42 21 2 ( Cell 105.485, 105.485, 62.445, 90, 90, 90)
Rmerge 6% to 2A

Solved MR using Phaser ( 1 mol in ASU) and built using COOT

Refined using Phenix to a 'final' Rwork 29.4% Rfree 31.7%

The maps look good but not great. Spurious peaks in difference maps.
Couldn't get R's down any further.

Xtriage suggested 'Overmerging pseudo-symmetric or twinned data'


Re-processed in P2 ( cell 105.299, 62.385, 105.548, 90, 90.021, 90)

Two molecules in ASU
Xtriage suggested the presence of  3 pseudo-merohedral twin operators
(one 4-fold and two 2-folds)
Refined with twin law h,-k,-l (one of the pseudo 2-folds) using Phenix
Rwork 26% Rfree 28% (no further building carried out at the moment)

Wondered if it was a higher space group so....

re-processed in P222 (again Rmerge 6% to 2A) and did MR with Phaser in
all the alternative space groups.

The highest LL-gain was for P 21 21 21 (cell 62.377 105.297 105.546 90
90 90)

I re-processed the data in P 21 21 21, all the systematic absences
looked 'absent'.

Xtriage suggested the presence of only one 4-fold Pseudo-merohedral twin
law h,l,-k, R obs 5.1%

Simulated annealing and ADP refinement in Phenix with twin law h,l,-k
gave Rwork 33% Rfree 40%.

Here are my questions...
1) Should I go for P2 due to the low R-factors and build from my good
model ( beta angle of 90.021 seems unlikely )
2) Should I try other possible SG ( HKL2000 suggested, primitive
tetragonal, primitive orthorhombic, C centred orthorhombic, and 
monoclinic).
3) I'm concerned that my free-R reflections will be screwed-up since I
have pseudo-merohedral twinning and some Rwork reflections may be the
same as Rfree ones. I have read that Phenix is rather clever and chooses
the highest observed symmetry to ovoid this problem but I can't see it
being reported in the log-file that, in the case of P2, pseudo P42 21 2
is used for Rfree calculations.

All help is much appreciated
Mike

-- 


Michael Gorman PhD
St. Vincent's Institute of Medical Research
9 Princes Street
Fitzroy
Victoria 3065
Australia
Tel : (03) 9288 2519