I agree with the general points that Mark makes. For the basic problem to find out whether the model quality (and R-values) is limited by the data quality, you can compare CCwork with CC* - if CCwork is significantly lower than CC*, then this means that either your model could in principle be improved (you have not yet found the right parameterization), or the data suffers from correlated systematic errors. So it would be good to try the XDS data processing.
best, Kay On Sat, 22 Feb 2014 23:15:33 +0100, Mark van Raaij <[email protected]> wrote: >As the excellent tips that you got indicate, lower R-factors can be obtained >by getting better data (better crystals, better data collection, better data >processing) or better fitting, i.e. refinement. In this respect, I am >impressed by the automatic data processing protocols now being implemented. >Also, the automatic local NCS refinement in REFMAC seems very good for our >recent structures. >But I would really want to make a general comment - not ALL structures can be >better than the average! There will always be structures with 5% higher >R/Rfree than the average in the same resolution range. Sometimes this will be >due to suboptimal refinement, but sometimes it may simply not be possible to >get better crystals and better data. Better not necessarily in term of >resolution, but in terms of disorders like you describe for your plate-shaped >crystals. >What I mean is that one should make all efforts to get better crystals and >data and refine structures as well as possible, but sometimes it may not be >possible to beat the average of the pdb and one should not get too hung up by >that. These structures should also be deposited and published. >On the other hand, these "rules" that R-factor should be a certain value at a >certain resolution, may lead to suboptimal refinement. For example the thought >"my R-factor is already better than the average" could be counterproductive >and lead people to stop refinement prematurely. >Sometimes a structure will have Rs better than the average for the resolution, >but still better refinement could lower it further and this should then be >done. I can think of an MR solution using a very homologous model that was >refined at higher resolution, structures with high NCS, or simply certain >rock-solid proteins... >Another popular one is (was?) that Rfree should always be below 30%, while >several important structures justifiably have Rfrees quite a bit higher >(others perhaps have not been refined enough). >So while comparing R/Rfree to the average of existing structures is useful, it >may not necessarily be a sign that a structure is "bad" if your Rs are 5 % >higher, not should your Rs being at or below the average be an excuse for >stopping refinement too early. >Fear that ones Rs are not low enough may even lead to certain forms of >cheating, for example not keeping the Rfree reflections truly free. > >On 22 Feb 2014, at 01:41, Chris Fage wrote: > >> Dear CCP4BB Users, >> >> I recently collected a number of datasets from plate-shaped crystals >> that diffracted to 1.9-2.0 angstroms and yielded very nice electron >> density maps. There is no major density unaccounted for by the model; >> however, I am unable to decrease Rwork and Rfree beyond ~0.25 and >> ~0.30, respectively. Probably due to the more 2-dimensional nature of >> my crystals, there is a range of phi angles in which the reflections >> are smeared, and I am wondering if the problem lies therein. >> >> I would be grateful if anyone could provide advice for improving my >> refinement statistics, as I was under the impression that the >> R-factors should be ~5% lower for the given resolution. >> >> A few more pieces of information: >> -Space group = P21, with 2 monomers per asymmetric unit; >> -Chi square = 1.0-1.5; >> -Rmerge = 0.10-0.15; >> -Data were processed in HKL2000 and refined in Refmac5 and/or phenix.refine; >> -PHENIX Xtriage does not detect twinning, but hints at possible weak >> translational pseudosymmetry; >> -I was previously able to grow one atypically thick crystal which >> diffracted to 1.65 angstroms with Rwork/Rfree at 0.18/0.22. >> Unfortunately, the completeness of the dataset was only ~90%. >> >> Regards, >> Chris > >Mark J van Raaij >Lab 20B >Dpto de Estructura de Macromoleculas >Centro Nacional de Biotecnologia - CSIC >c/Darwin 3 >E-28049 Madrid, Spain >tel. (+34) 91 585 4616 >http://www.cnb.csic.es/~mjvanraaij > >
