We had recently a similar case. Indexed in P2x2x2x but truly was P21 with variable amount of twin fraction depending on the dataset & beamline between 9% and 40%. We were able to solve it using phenix.ensembler using all available known structures plus our various homology models. We still have not gone back and done a postmortem analysis of why we failed with a highly similar homolog to find a solution in the first place with either Phaser, Molrep or Phenix using single models. It might help to process your data in XDS to take care of the overlaps a bit better in case you have not done so already.
Here's a link <advertisement on> Hain et al. 2012 http://www.ncbi.nlm.nih.gov/pubmed/22982544 </advertisement off>. Jürgen ...................... Jürgen Bosch Johns Hopkins University Bloomberg School of Public Health Department of Biochemistry & Molecular Biology Johns Hopkins Malaria Research Institute 615 North Wolfe Street, W8708 Baltimore, MD 21205 Office: +1-410-614-4742 Lab: +1-410-614-4894 Fax: +1-410-955-2926 http://lupo.jhsph.edu On Oct 26, 2012, at 8:27 AM, Seijo, Jose A. Cuesta wrote: Hi all, I am dealing with a molecular replacement problem for a 60KDa protein composed of 2 rigid domains joined by a flexible linker which can move relative to each other. Sequence identity for my best model is 46% evenly spread, so in principle this should be a tractable problem. Then the problems start to pile up: a) The unit cell is 56.7Å, 288.5Å, 69.4Å, 90 93.5, 90. Spacegroup P21. Rmerge 12% to 2.4Å. The data also merges relatively well (Rmerge 16%) in P222 with the same a, c and b axes, now of course in that order. In the P21 case, that corresponds to 4 monomers in the asymmetric unit with a solvent content of approx. 50%, giving me 8 domains to find if I separate them. b) The 288 axis means that my data show some overlap in almost all orientations (might be corrected in the future with new datasets), so that my low resolution data are likely unreliable. c) Intensity distributions suggest twinning in either point groups. Actually, they are beyond the perfect twinning case, which I attribute to the reflection overlaps making the strong reflections weaker (integration box too small) and the small stronger (from tails of adjacent strong ones). Of course the latest would mean that the twin fraction estimation is unreliable, but all moments, etc show perfect twin statistics, so I am assuming that there is indeed perfect twinning of some sort. So, the question is, what is the best strategy to deal with this many (4 or 8) body / noisy / twinned problem? I am trying EPMR with many bodies, but I suspect the twinning would throw it out of the right track, and one domain seems to be too little of the diffracting matter to show any sort of discriminations between solutions and non-solutions if do the usual serial searches. I plan to let autotracing programs be the judge of success, but I am not sure of how well those can deal with twinning. Can Arp-Warp use twinned data? Thanks in advance for any tips. Jose. ======================================== Jose Antonio Cuesta-Seijo, PhD Carlsberg Laboratory Gamle Carlsberg Vej 10 DK-1799 Copenhagen V Denmark Tlf +45 3327 5332 Email josea.cuesta.se...@carlsberglab.dk<mailto:josea.cuesta.se...@carlsberglab.dk> ========================================
