I think your idea was discussed herein before, but that the consensus was that CC1/2 does not change much when you scramble and re-calculate.
What one needs is a good way to determine at what point in the CC1/2 curve the data stops being useful in obtaining the best possible model, and it seems that this point may depend on individual cases. In a way, there is still some information in reflections with I/sigma approaching one, since even those measurements limit the possible values of I if only slightly, i.e., the true value of I can't be greater than 1000 or something. I think that if crystallography still exists in 20 years, we will be refining directly against pixel/voxel/roxel(?) intensities. JPK -----Original Message----- From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of Takanori Nakane Sent: Thursday, July 02, 2015 8:53 PM To: CCP4BB@JISCMAIL.AC.UK Subject: Re: [ccp4bb] paired refinement Hi all, I have read recent SERCA paper on IUCrJ and found their discussion interesting. "Structural studies of P-type ATPase–ligand complexes using an X-ray free-electron laser" by Maike Bublitz et al. http://journals.iucr.org/m/issues/2015/04/00/jt5009/index.html In addition to CC1/2 and paired refinement, they scrambled indices of reflections and showed Rfree indeed went up. If we repeat permutations many times and calculate CC1/2 every time, we can obtain empirical distribution of CC1/2 under null hypothesis and thus calculate p-value. This is called permutation test in statistics. I am wondering how this p-value compares with p-values calculated by Student's t-test. Since t-test does not take the effect of multiplicity into the account, I think the permutation test gives more accurate value. (That is, CC1/2 calculated from a dataset with multiplicity 30 should be more accurate and "reliable" than that calculated from a dataset with multiplicity 2, but this difference is ignored in t-test.) Did anyone try this? Best regards, Takanori Nakane On 2015年07月03日 04:41, Tim Gruene wrote: > Hi Robbie, > > I have been wondering how much information would be present in a > weighted CC1/2 with weights from the ML refinement program. > > As I understand the concept behind paired refinement, one can use much > higher resolution data in refinement than you would expect from the > (classical) scaling statistics, because refinement programs > down-weight the outliers. > > Maybe a weighted CC1/2 would be a convenient short cut the more time > consuming paired refinemement? How easily could the calculation of a > weighted CC1/2 be implemented in refmac? Refmac would have to merge > the data internally, but that's just a simple formula to implement. > > Best, > Tim > > On 07/02/2015 08:28 PM, Robbie Joosten wrote: >> Hi Jacob, >> >> You need unmerged data to calculate cc1/2. That's not the sort of data you >> get from the PDB. But anyway, we have a fairly simple automated test that >> we can use on a case-by-case basis. I would argue that that is nicer than a >> empirical cut-off that may or may not be correct for you case. >> >> Cheers, >> Robbie >> >> Sent with my Windows Phone >> ________________________________ >> Van: Keller, Jacob<mailto:kell...@janelia.hhmi.org> >> Verzonden: 2-7-2015 20:12 >> Aan: CCP4BB@JISCMAIL.AC.UK<mailto:CCP4BB@JISCMAIL.AC.UK> >> Onderwerp: Re: [ccp4bb] paired refinement >> >> Well, in that case, one could simply look at the plot of CC1/2 versus >> resolution and see the step up to one, conclude something was off. >> >> I wonder whether PDB REDO was able to get some empirically-determined values >> for CC1/2 cutoffs by comparing paired refinement versus CC1/2 versus other >> parameters? >> >> JPK >> >> -----Original Message----- >> From: CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] On Behalf Of >> Dale Tronrud >> Sent: Thursday, July 02, 2015 1:46 PM >> To: CCP4BB@JISCMAIL.AC.UK >> Subject: Re: [ccp4bb] paired refinement >> >> While I was puzzling over an entry in the PDB some years ago >> (since >> obsoleted) I noticed that all the high resolution amplitudes were equal to >> 11.0! This was before CC1/2 but for this structure it would have been equal >> to one, and yet the outer data were useless. A practical test like paired >> refinement can't be fooled in this way. >> >> Dale Tronrud >> >> On 7/2/2015 10:25 AM, Edward A. Berry wrote: >>> My take on this- >>> No one has been willing to specify a cutoff (and probably there is >>> no rigorous way to mathematically define the cutoff) and say "If CC* >>> (or CCfree or >>> whatever) is below X >>> then it will not improve your structure, if above X then it will". >>> Probably depends >>> among other things on how strong the lower resolution data is, how >>> good the structure is without the added data. >>> On the other hand in paired refinement, if adding the data improves >>> the structure as measured by Rfree in a zone excluding the added >>> data, then it is hard to deny that that data are worth including. >>> >>> eab >>> >>> On 07/02/2015 12:52 PM, Keller, Jacob wrote: >>>> Wasn’t all of this put to bed through the implementation of CC measures? >>>> >>>> JPK >>>> >>>> *From:*CCP4 bulletin board [mailto:CCP4BB@JISCMAIL.AC.UK] *On >>>> Behalf Of *Robbie Joosten >>>> *Sent:* Thursday, July 02, 2015 12:46 PM >>>> *To:* CCP4BB@JISCMAIL.AC.UK >>>> *Subject:* Re: [ccp4bb] paired refinement >>>> >>>> But it is not the R-free of the shell here. In paired refinement >>>> you take the R-free of the reflections outside the shell. >>>> >>>> Cheers, >>>> Robbie >>>> >>>> Sent with my Windows Phone >>>> >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> ------------------------------------------------------------------- >>>> -- >>>> --------------------- >> --- >>>> >>>> >>>> *Van: *Edward A. Berry <mailto:ber...@upstate.edu> >>>> *Verzonden: *2-7-2015 18:43 >>>> *Aan: *CCP4BB@JISCMAIL.AC.UK <mailto:CCP4BB@JISCMAIL.AC.UK> >>>> *Onderwerp: *Re: [ccp4bb] paired refinement >>>> >>>> Another criterion for cutoff, also requiring the structure to be >>>> solved, is the agreement between data and structure, e.g. Rfree or CCfree. >>>> I think it is very unlikely that you could get Rfree =.2493 in a >>>> shell which contains only noise. So I would suggest doing paired >>>> refinement to 2.2 and 2.1 A (if the data is available). >>>> >>>> On 07/01/2015 07:15 PM, Eric Karg wrote: >>>> > Hi all, >>>> > >>>> > I have a dataset processed in XDS to 2.3 A (based on CC1/2). >>>> I'm trying to do "paired refinement" to determine the optimal >>>> resolution cutoff. Here is what I get at different resolutions set in >>>> Phenix: >>>> > >>>> > Final Rfree/Rwork: >>>> > 2.7—> 0.2498/0.2027 >>>> > 2.6—> 0.2519/0.2009 >>>> > 2.5—> 0.2567/0.2025 >>>> > 2.4 —> 0.2481/0.2042 >>>> > 2.3 —> 0.2493/0.2075 >>>> > >>>> > The geometry of all output structures are similar. >>>> > >>>> > 1. What is the high resolution cutoff based on these data? I >>>> know that Rfree/Rwork at different resolution should not be >>>> compared, but is there a simple way to do the test as described in >>>> the K&D 2012 Science paper using Phenix GUI? >>>> > >>>> > 2. For refining a structure at a lower resolution (lower than >>>> the initial dataset), do I simply set the resolution limit in the >>>> refinement or I need to reprocess the data starting from the images? >>>> Do I need to do anything with Rfree flags? Based on the discussions >>>> on this forum I know I should deposit the highest resolution >>>> dataset but my question is about the mtz file which will be used for >>>> refinement. >>>> > >>>> > Thank you very much for your help! >>>> > >>>> >> >