[phenixbb] Pseudo-merohedral twinning refinement in C2, when beta = 94 degrees

[John Pak]

Hi all,

I currently have a dataset that scales well in C2.  (a = 224,b = 129,c = 
392. beta= 94.5 degrees).  Previously, I had thought that these crystals 
were in P6322 (posted this a few months ago), but I've since grown 
better crystals, where the beta angle is now 94.5, and not close to 
90.0.  I'm reasonably confident with the scaling - however, it has 
always been really difficult to index these crystals in HKL2000.

Like the previous crystals, this new C2 dataset looks like it might be 
perfectly twinned, based on Xtriage's L test.

I've solved the structure by MR, and refined it to 3.0 angstroms to 
R=30%, Rfree=33%, but I am running out of things to fix.  I'd like to 
perform twinning refinement, based on this Xtriage L test result.

The problem is that Xtriage cannot find any pseudomerohedral twinning 
laws, likely because the beta angle is too far from 90 degrees??  For 
kicks and giggles, I started to enter arbitrary beta angles in Xtriage, 
and at a beta angle below 93.4, Xtriage can find two twinning laws 
(closer to 90 degrees, Xtriage can find 5 twin laws).  Using 
"-1/2*h+3/2*k, 1/2*h+1/2*k, -l" in Phenix refine (the beta angle is 
still 94.5 degrees, I haven't changed this) results in the R and Rfree 
dropping a lot to 25.5 and 28.6, respectively.  Used the same random 
seed as previous runs to generate the test set from the scalepack file, 
with the "use lattice symmetry to generate test set" flag enabled.  The 
maps look a lot noisier.

So my questions are:

1. Is this twinning refinement valid?  I'm really thinking no - since 
the maps didn't improve.   Suppose the maps didn't get worse though - 
would this twinning refinement then be valid?

2. How do I go about reindexing this dataset in I2? i.e. to see if the 
beta angle with this unit cell selection is closer to 90 degrees.  Would 
this even help, with respect to pseudomerohedral twinning refinement?


Thanks!