Hi JK,

we recently worked on a crystal with similar symmetry and I would recommend you to read the two papers listed below. If you visually do see the "weak" reflections, the space group is primitive orthorhombic (and not body-centered orthorhombic). The strong Patterson peak at ~0.5, ~0.5, ~0.5 results from pseudo-translational non-crystallographic symmetry (tNCS), as you correctly pointed out. As a consequence the intensity distribution is different compared to a "regular" crystal without tNCS, which affects molecular replacement and also coordinate refinement.

The tNCS could obscure the determination whether your crystallographic two-fold symmetry axes are screw axis or not, thus your space group could in principle also be P22121, P21212, P21221, P2122, P2212, P2221 or P222. This is nicely described in the first reference.

In light of the tNCS, you may also want to give Phaser a try for molecular replacement, I would try all eight primitive orthorhombic space groups (keyword: SGALTERNATIVE SELECT ALL).

Twinning could in theory be possible (e.g. a P2 or P21 twin with the unique beta cell angle very close to 90 degree [could be any of the three angles]), but difficult to detect together with the tNCS.

Best, Simon

J. A. Sundlov, A. M. Gulick, Structure determination of the functional domain interaction of a chimeric nonribosomal peptide synthetase from a challenging crystal with noncrystallographic translational symmetry. Acta Crystallogr D Biol Crystallogr 69, 1482-1492 (2013)10.1107/S0907444913009372).

R. J. Read, P. D. Adams, A. J. McCoy, Intensity statistics in the presence of translational noncrystallographic symmetry. Acta Crystallogr D Biol Crystallogr 69, 176-183 (2013); published online EpubFeb (10.1107/S0907444912045374).

On Tue, Feb 17, 2015 at 7:19 PM, Jayakrishnan Nandakumar User <[log in to unmask]> wrote:

We have data at ~3.2 A for a crystal that looked clearly like primitive orthorhombic while indexing in HKL2000 (or Mosflm). Based on the systematic absences I called it P212121 during scaling. However, we are having some unexpected issues with finding a Mol. Rep. solution using a model that is present as is in the crystal (along with regions that are absent in the model). There seems to be a clear non-origin Patterson peak at ~0.5, ~0.5, ~0.5. In agreement with this, when I use Molrep to find two molecules of the model in the asymmetric unit, the two solutions are derived from the same rotation function but with translation solutions displaced exactly 0.5, 0.5, 0.5 fractional coordinates from each other. However, the maps after refinement don't look impressive at all.

I looked at h+k+l = 2n+1 reflections. Visual inspection suggests that several of these are weaker than the corresponding h+k+l = 2n reflections, but there are certainly exceptions. I have not done a systematic calculation of what the average I and sigma are for h+k+l = 2n+1 and h+k+l = 2n are.

Does this smell like a wrong space group assignment? Twinning? Or am I missing something very trivial here? Or is this a pseudo-translation we have to live with?

Any kind of advice will be very helpful. I am happy to provide more information if you tell me what you need.

Thanks a ton in advance.
Best,
JK
P.S. Matthew's coefficient indicates that there could easily exist up to 6 monomers in the asymmetric unit (4 monomers if 70% of the crystal is solvent).