Having run foul of the "R-factor police" on a number of occasions, but having ultimately prevailed, I would like to offer the following opinion:
Intensity measurements in the weak outer shells are as valid as any other and, if they are correctly processed and properly assessed for their statistical robustness, their merged values are as valid as any other. These data will contribute little to the refinement, but they will still contribute, so why ignore that contribution given the sophistication of modern refinement software. Nevertheless, in light of the above, I would suggest that we now avoid using statements such as "The structure of protein X determined at 2.0 Angstoms resolution", instead replacing them with now more accurate statements such as "The structure of Protein X determined using data to 2.0A resolution" and then let the maps speak for themselves! with best wishes Mike Lawrence Associate Professor and WEHI Fellow Division of Structural Biology Walter and Eliza Hall Institute of Medical Research 1G Royal Parade, Parkville Victoria 3052, AUSTRALIA On 11/10/2013, at 9:44 AM, Jim Pflugrath wrote: > Please tell me why Rpim should be looked at. Cannot one have meaningless > data and have lots of multiplicity to drive Rpim lower without any real > benefit? Under what conditions is Rpim useful? > > And suppose one looks at <I/sigI> (and not <I>/<sigI>) and CC1/2. What of it? > > And let me write what Phil wrote in a slightly different way: > Please explain how you think that adding the resolution from 2.6 A to 2.45 A > will improve your model. > > Sorry, but maybe it is too soon after the last CC1/2 discussion to raise > these points, but I am truly interested in various opinions about all this. > > ________________________________________ > From: CCP4 bulletin board [[email protected]] on behalf of Gerard > Bricogne [[email protected]] > Sent: Thursday, October 10, 2013 5:28 PM > To: [email protected] > Subject: Re: [ccp4bb] how to cut back resolution of a well-refined model > > Dear Yafang, > > Is it the case that you collected these data on a Pilatus detector, > using relatively low exposure and high multiplicity? These types of datasets > always give what looks like alarmingly high values of R-merge, and many > people who are set in their ways (like so many reviewers still are) tend to > conclude that the alarm is about the data being bad, whereas it is about > Rmerge being a terrible statistic in these situations. The Rpim statistic, > on the other hand, is the one to look at if you want an R-like quantity, and > it is well behaved in this regime. Of course, look at CC1/2 as well, and > I/sigI as you did. > > > With best wishes, > > Gerard. > > -- > On Thu, Oct 10, 2013 at 04:57:20PM -0400, Yafang Chen wrote: >> Hi All, >> >> I have a structure at 2.45A which has been well refined. However, since the >> R-merge at the last shell is above 1 (although I/sigmaI at the last shell >> is more than 2), we now decide to cut back the resolution to about 2.6A. Is >> there a way to do this based on the well-refined model instead of doing the >> MR and refinement all over again? Thank you so much for your help! >> >> Best, >> Yafang >> >> -- >> Yafang Chen >> >> Graduate Research Assistant >> Mesecar Lab >> Department of Biological Sciences >> Purdue University >> Hockmeyer Hall of Structural Biology >> 240 S. Martin Jischke Drive >> West Lafayette, IN 47907 > > -- > > =============================================================== > * * > * Gerard Bricogne [email protected] * > * * > * Global Phasing Ltd. * > * Sheraton House, Castle Park Tel: +44-(0)1223-353033 * > * Cambridge CB3 0AX, UK Fax: +44-(0)1223-366889 * > * * > =============================================================== ______________________________________________________________________ The information in this email is confidential and intended solely for the addressee. You must not disclose, forward, print or use it without the permission of the sender. ______________________________________________________________________
