Hi Florian,
When your pseudo-translation (native Patterson peak) is close to (0.5,0.5,0.5) even in p22121 you will have have most of reflections with
h+k+l=2n+1
measured as weak. In your lower resolution datasets 2.6 and 3.0 A non-weak reflections can be lost in noise and your group assignment
may go wrong, and all of your crystals may be P22121.
Having said this, polymorphism is quite common.
In a recent publication Acta Crystallogr D Biol Crystallogr. 2014 Sep 1;70(Pt 9):2430-43, Lebedev & Isupov
we describe a case where two crystals had similar cell parameters in different space groups
a = 119.2, b = 192.5, c = 77.3 P212121
and
a = 112.0, b = 192.2, c = 76.7 P21212.
Both structures could be refined in the other space group at 1.8A resolution, but resulting R-frees were 8-15 % higher in the wrong space group.
Both space group appeared as a result of symmetry breakdown in a supergroup A2122 (reindexed C2221).
P212121 structure could also be refined in the supergroup with FreeR only 3 % higher than in the real space group.
However as half of reflections had to be thrown out to reindex to A2122 even if FreeR would be equal in the supergroup and P212121
I would consider the latter correct.
I would recommend to solve and refine your low resolution structures in all possible space groups and choose the one with better refinement statistics,
preferring not centered one in case of equal Rfrees.
Quality of density maps could be criteria in borderline cases.
Cheers,
Misha
________________________________________
From: CCP4 bulletin board [[log in to unmask]] on behalf of Florian Schmitzberger [[log in to unmask]]
Sent: 13 October 2014 09:47
To: [log in to unmask]
Subject: [ccp4bb] I222 - P22121 space-group ambiguity
Hi everybody,
I collected a number of X-ray data sets from crystals originating from the same cryst. drop. I solved the initial structure in P22121 space group by MR with Phaser locating two molecules (data to ~ 2.1 Angstr.); refined R/Rfree: 0.213/0.244.
Processing of some of the other data sets with XDS/Aimless is consistent with I222 space group (resolution ~ 2.6 Ang.). I can locate one molecule. The unit-cell dimensions for I222 and the initial P22121 space groups for two of the data sets are:
I222: a=87.8 b=101.18 c=123.63; P22121: a=93.34 b=105.47 c=122.98;
I superposed the molecule in I222 onto one of the two located for the initially solved P22121; the orientation of the NCS-related molecule in P22121 differs from the crystallographic-symmetry related one in I222. Trying to solve this P22121 data set in I222 with MR, does not result in high Z scores, and maps do not look good.
Some of the data sets that process in I222 to ~ 3 Angstr., I can also solve in P22121, locating two molecules (differences may not be that clear in this case, since the resolution is lower).
Some other data sets process in P22121 with Aimless; with a substantial off-origin Patterson peak, indicating translational NCS. For these, Phaser positions two molecules that are related by crystallographic translational NCS. These two molecules are crystallographic-symmetry related in the original P22121 data set. I can also solve these data sets in I222 space group, with the overall Z score higher than for the P22121 data.
I am uncertain, what the ‘true’ space group for some of my data sets is. Could it be that for data that process in P22121, but can be solved in I222, reflections that would indicate I222 space group were not collected? Alternatively, perhaps what I am seeing is that there is a (gradual) transition of the crystal lattice (between P22121 and I222 or vice versa), caused by variation in crystal handling/cooling or exposure to X-rays.
It’s relevant to me, because in P22121 space group, a region of the molecule that is of biological interest makes NCS-related crystal contacts that are crystallographic-symmetry related in I222.
Has anybody observed similar cases? I would appreciate comments.
Cheers,
Florian
|