Hi Edward,
I and also others in the bulletin boards have had cases where the protein would build e.g. tetramers with internal 222 symmetry, which would pack with the 2-folds parallel to the crystallographic 2-folds with say one 2-fold a little shifted from the crystallographic position. In these cases, the internal 222 symmetry overwhelmed the contribution of the small shift and all processing programs, including Pointless would insist that the crystal was P2x2x2x. Also the Rfactors (Rsym, Rmeas) would be the same whether processing in P222 or P2.
However, the only way to pack the proteins correctly in the unit cell, would be to process in P2 or even in P1 and run molecular replacement in P2 or P1 and than reconstruct the most likely "true" space group, for which now the program Zanuda exists. Since in these cases the diffraction patterns would have "perfect" (within experimental error) P222 symmetry, I think it would be valid to expand the P222 data to P2 of P1. Of course collecting the full data set is still the best way to go.
The asymmetry is generated by the way the molecular replacement program places the molecules in the asymmetric unit and will, as you mention, only be in the phases. Expanding to P1 and running molecular replacement in P1 would test all possibilities at once, but the signal would be weaker. A more cautious approach would be to process in the three possible P2 spacegroups (choosing the 2-fold along a, b or c) and run molecular replacement for each option. It may also be that your problem lies elsewhere.
Best,
Herman
-----Original Message-----
From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of Edward A. Berry
Sent: Monday, April 29, 2013 4:14 PM
To: [log in to unmask]
Subject: Re: [ccp4bb] refinement hanging--what am I missing?
[log in to unmask] wrote:
> I would process or expand the data to P1, also expand your pdb file to P1 and refine in P1 to see what happens. I would also run Phaser or some other molecular replacement program on the P1 data to see what comes out.
process or expand? I don't understand how there can really be a choice here.
If the data is expanded from higher symmetry, it will precisely have that higher symmetry, i.e. all reflections that were equivalent in the higher symmetry will be identical. Any information about asymmetry was lost when the data were merged in the higher symmetry space group. Won't refinement by minimizing a target function have exactly the same solution?
Or not exactly, because the refinement program can introduce asymmetry by allowing different phases for the equivalent reflections.
But where is the information to generate that asymmetry coming from?
There is probably something I'm missing here, but i would definitely reprocess in the lower symmetry if the rotation angle was sufficient to give good completeness.
eab
|