Thanks Randy, so from your reply it seems that cutoff is differently treated. And if I interpret your email correctly it is better to provide Phaser with a truncated versus a full data set. I tried both cases, but I had assumed that if you restrict the resolution within Phaser it would be the same as if you give Phaser to begin with a truncated dataset. Perhaps it would be a good idea in th automatic routine to first check if the user wants to truncate the data and then run the automatic scoring analysis with that value ?
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<tel:%2B1-410-614-4742> Lab: +1-410-614-4894<tel:%2B1-410-614-4894> Fax: +1-410-955-2926<tel:%2B1-410-955-2926> http://lupo.jhsph.edu On Oct 1, 2014, at 5:15 AM, Randy Read <[email protected]<mailto:[email protected]>> wrote: Hi Jurgen, You could send me a logfile off-list, and maybe I would spot something in there. We’ve put some effort into putting more intelligence into the Phaser search, so that it adapts to the initial perceived difficulty of the problem in setting the initial parameters, and then adapts to indicators of success or failure during the search. Much of the time this works very well, but there’s obviously room for improvement. For one thing, it appears that we’re frequently too optimistic about how good the model will be and how easy the search will be. One question: when you say that you cut at 2.5A for MR, do you do that by setting the resolution within the Phaser run, or do you have an MTZ file with only data to 2.5A? If the former, then you’re over-riding some of Phaser’s automation, which will choose the initial resolution limit based on the perceived difficulty of the problem (a function of model completeness, expected RMS error of the model, and the number of reflections to different resolution limits). It’s this initial automated choice that can go wrong if we’re too optimistic, because then a clear solution isn’t found, and then Phaser repeats the search with data to the full resolution, which can take longer than just choosing an intermediate resolution from the start. Anyway, if the problem is expected to be easy but turns out to be difficult, this implies that some of the information used to decide it should be easy is wrong or too optimistic. One top possibility is that the model is not as good as expected, e.g. because of conformational changes. If there’s a potential hinge-bending motion, then you’re usually better off searching with separate domains. If the change is something that can’t be described with rigid-body motions, then it would be better to increase the expected RMS error from what Phaser deduces from the sequence identity. Increasing the RMSD by 10-20% would be a good first bet in such a case. The other top possibility is that the space group is wrong. Is there any ambiguity in the space group? In particular, do any of the twinning tests indicate that the data may be twinned (which can lead to choosing too high symmetry)? Best wishes, Randy On 1 Oct 2014, at 03:02, Jurgen Bosch <[email protected]<mailto:[email protected]>> wrote: Dear BB, or in particular Phaser developers :-) This must be part of British humor right (or was that the Canadian influence Randy) ? eLLG indicates that placement of ensemble "ensemble_1" will be straightforward The data are sufficient to exceed the eLLG target The search space is finite 143 x 143 x 80 Å with 3 or 4 molecules per asu, but Phaser has been burning CPU cycles quite a bit, we are approaching 24h by now. Data extends to 1.7 Å - for MR we cut at 2.5 Å. I can hear Garib, yes, Molrep was done in few minutes but I’m not super convinced about the solution either. Same with BALBES and MrBump (which took a few more minutes, actually also days for MrBump) The space group appears to be P63 22 as judged by XDS, pointless, xtriage, however the crystals are split (at least in some areas it’s visible) but I thought XDS would take care of these “aliens” and eliminate them mostly. My graduate student, Lauren, found a nifty program called DIALS that we wish to explore further to rescue the “nice” data we have and hopefully solve the structure. Lower symmetry space groups were tried down to P21 with increasing number of molecules per asu and applying twin laws if necessary. The minor problem with the twins is how do I really know that it is a higher symmetry space group and not a 50% twin in a lower symmetry ? Rwork/Rfree in this particular case do not seem to help at all for distinguishing between the solutions. Maps looks sort of right but R-factors are not reflecting what you see on the screen. We appreciate any suggestions and ideas what else we could do. Thanks, 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<tel:%2B1-410-614-4742> Lab: +1-410-614-4894<tel:%2B1-410-614-4894> Fax: +1-410-955-2926<tel:%2B1-410-955-2926> http://lupo.jhsph.edu<http://lupo.jhsph.edu/> ------ Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: + 44 1223 336500 Wellcome Trust/MRC Building Fax: + 44 1223 336827 Hills Road E-mail: [email protected]<mailto:[email protected]> Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk
