[phenixbb] Need help in refining my twinned P6322(apparent) dataset

[Phil Jeffrey]

John,

The test here would be "which maps looks better" and if there isn't any 
obvious interpretable difference density for a missing/modified part of 
the structure you might generate some by leaving out a well-ordered 
element of the structure prior to the first pass through refinement in 
each of your P321 and P63 cases.

Personally I would rebuild both P321 and P63 structures at the same 
time, and see if any meaningful differences manifest.

That having been said, I want to raise the unfortunate possibility that 
you're really in P3 and doubly twinned.  As far as I know phenix.refine 
does not handle that so it's difficult to test for.  However I did have 
one case that appeared to fit within that scheme (also in trigonal).

If you didn't generate your Free R flags within Phenix itself (which 
typically handles this correctly), you need to make sure that they obey 
point group 622 symmetry even in lower symmetry trials, to minimize 
R-free biasing due to NCS and twinning.

Phil Jeffrey
Princeton



On 3/15/11 2:55 AM, John Pak wrote:
> Hey Phenix community,
>
> Here are the details of my current situation - any assistance in the
> matter would be greatly appreciated as I've been struggling mightily for
> the past month trying to get this structure refined (to no avail).
>
> Details:
>
> -3.0 angstrom dataset from a membrane protein crystal
> -reduced in a P3 Bravis lattice using HKL2000
> -Pointless and Xtriage suggest P622 symmetry
> -Solved structure via MR - ran Phaser in all P622 based space groups,
> found 1 convincing solution in P6322 (LLG ~110) which makes biological
> sense (1 monomer in ASU that lies on the screw axis to recapitulate the
> biological trimer)
> -31% sequence identical search molecule
> -no pseudotranslation (via Xtriage)
>
> Xtriage suggests that the P6322 dataset is nearly perfectly twinned
> (twin fraction = 0.49). Mindful of the fact that there are no twin
> operators for P6322, I'm thinking that the true space group is P321,
> P312 or P63 and I've tried a number of things listed below:
>
> 1. Refined in P6322 (simulated annealing+individual B). Missing side
> chains and incorrectly modelled loops apparent in maps. Start R=0.49,
> Rfree=0.5. Refinement stalls at R=0.41, Rfree=0.45 after building.
>
> 2. Ran Phaser in P312 (using unrefined model) - no solution.
>
> 3. Ran Phaser in P321 (using unrefined model). One nice solution (LLG
> ~220) - lattice looks very similar to P6322 crystal packing. Twin
> refinement in phenix using operator suggested by Xtriage (simulated
> annealing + individual B) results in R=0.38, Rfree=0.41 (no rebuilding
> performed).
>
> 4. Ran Phaser in P63 (using unrefined model). Three solutions (LLGs all
> ~210) - solution #3 looks very similar to P6322 crystal packing. Twin
> refinement in phenix using operator suggested by Xtriage of solution #3
> gives R=0.37, Rfree=0.40 (no rebuilding performed). Solutions #1 and #2
> do not give packing arrangements that look like the P6322 solution. Twin
> refinement of Sol 1 and 2 give higher R factors.
>
> So I'm at a loss as to which space group my crystal truly belongs to,
> and where I should be focusing my refinement. The maps all look similar
> before and after DM.
>
> Thanks for any assistance or recommendations that you might have!