Hi Wenhua You didn't say what resolution your structure is: that's critical to deciding how significant is the R-factor difference (or more strictly the ratio). There are other factors such as solvent content: generally anything which affects the obs/param ratio (such as NCS restraints) will also affect what value of Rfree you would expect to get given a value of Rwork, assuming purely random errors and a structure refined to convergence. Note that this doesn't tell you directly whether there's overfitting: that's more a matter of comparing results from different parameterisations & weighting schemes.
The value of Rwork itself will affect the Rfree-Rwork difference: higher values of Rwork tend to give higher Rfree-Rwork, other things being equal (using the Rfree/Rwork ratio instead of the difference normalises the difference and removes this dependence on Rwork). For example for 3A resolution or lower, a difference of 0.1 may not be indicative of a problem (depending as I said on the value of Rwork). At 1A it almost certainly would indicate a problem since at that resolution you would normally expect a ratio of say around 1.2 ( or a difference of around 0.02 for Rwork = 0.1). At 3A or lower resolution you expect a ratio of say 1.4 or more so you could easily get a difference of 0.1 for Rwork ~= 0.25. For anything in between these extremes we need a bit more info! Cheers -- Ian On 26 November 2013 08:49, Wenhua Zhang <[email protected]> wrote: > Good Morning to ALL, > > I encountered large gaps between R-factor and R-free, eg. 10% and 11%, > during refining several structures with either Refmac or Phenix.refine. > The data were reduced and scaled at resolution with I/SIGMA higher than > 2.3. > The model fits and density pretty well despite a few side chains missing > density. > > I may ask for the reasons that result in such large difference between R > and R free and some ways to examine and improve the refinement. > > Thanks in advance > > Wenhua > > Univ. Paris-Sud XI, Orsay > > > On 11/25/2013 6:03 PM, Beatrice Vallone wrote: > > Please check the announcement for 2014 ESRF Users' Meeting & Associated > Structural Biology Workshop. > > Invitation to the 2014 ESRF Users' Meeting & Associated Workshops to meet > and discuss with fellow users, get inspired, establish new scientific > collaborations and learn about future opportunities offered by the ESRF for > your research. > > Workshop on* "Structural Biology at ESRF: Past, Present & Future > <http://www.esrf.eu/home/events/conferences/um2014/structural-biology-workshops.html>"* > 3 - 5 February > > The Plenary Meeting will feature: > > - 5 plenary lectures by world leading scientists covering a large > spectrum of ESRF activities; > - a presentation on the status of the facility and on the progress of > the Upgrade Programme; > - a talk by the winner of the prestigious "ESRF Young Scientist Award"; > - a poster session; > - a banquet, with awards to the ESRF Young Scientist Award and for the > best poster > > Programs for the Meeting and Workshops are available at: > > http://www.esrf.eu/home/events/conferences/um2014.html > > Regards to all, > > Beatrice Vallone > ESRF User Organization Committee > > > > >
