Dear Mike, I'd say: Initially, yes. You should try to copy the R-free set from the old dataset to the new one. Assuming the reason to switch from one set to the other is higher resolution data in the new set, you should not forget to check that you also have enough R-free reflections in the higher resolution bins.
The whole 'is my R-free set biased'-discussion comes around every so often at the CCP4BB and IMO there is consensus about this: - Yes, your set will be biased. - Yes, you should keep your original R-free set (whenever you can). - If you cannot keep the R-free set, extra precautions have to be taken. There is still discussion on how bad the bias is and what should be done if you cannot transfer the R-free set. Here are some suggestions: - Give the coordinates random shifts (how much?) - Do some (brutal) SA refinement - Reset all B-factors to a single value - Do extensive cartesian-space refinement - Combine the above Anyway, the more you change you model, the weaker you bias becomes. Typically, you R-free will be suspiciously low in the beginning and it will go up (a lot) when you do any of the above. Once it levels out, you can continue model building again. If you (or the referees) still have doubts about R-free bias, consider using the R/R-free ratio test discussed in "Tickle IJ, Laskowski RA, Moss DS. R-free and the R-free ratio. I. Derivation of expected values of cross-validation residuals used in macromolecular least-squares refinement. Acta Cryst. 1998; D54:547-557" HTH, Robbie Joosten Netherlands Cancer Institute Amsterdam, The Netherlands ---------------------------------------- > Date: Wed, 23 Sep 2009 22:37:17 -0700 > From: det...@uoxray.uoregon.edu > Subject: Re: [ccp4bb] Rfree in similar data set > To: CCP4BB@JISCMAIL.AC.UK > > Yes > > Dale Tronrud > > Mike England wrote: > > Hi all, > > > > I will appreciate your comments on the following case: > > > > I have two datasets from the same or identical crystals. Initially, I > > refine a structure against the first data set and later on switch to > > another dataset for further refinements. > > Do you think, my Rfree will be biased as Rfree reflections in second > > dataset may be in fact Rwork reflections in previous datasets ? > > > > Thanks in advance, > > > > Mike
