*** For details on how to be removed from this list visit the *** *** CCP4 home page http://www.ccp4.ac.uk ***
Mitch Miller brought to my attention two excellent papers from Ian Tickle and coworkers on the subject of expected Rfree/R ratio and how it depends on resolution and method of refinement. Note, in particular, Fig. 1 of each paper. Here they are: Rfree and the Rfree Ratio. I. Derivation of Expected Values of Cross-Validation Residuals Used in Macromolecular Least-Squares Refinement. Tickle IJ, Laskowski RA and Moss DS Acta Cryst. (1998). D54, 547-557 http://dx.doi.org/10.1107/S0907444997013875 Rfree and the Rfree ratio. II. Calculation of the expected values and variances of cross-validation statistics in macromolecular least-squares refinement. Tickle IJ, Laskowski RA and Moss DS Acta Cryst. (2000). D56, 442-450 http://dx.doi.org/10.1107/S0907444999016868 > From: <[EMAIL PROTECTED]> > Date: Wed, 05 Jul 2006 14:23:39 -0500 > To: <[email protected]> > Subject: Re: [ccp4bb]: Re: Re: [ccp4bb]: > > *** For details on how to be removed from this list visit the *** > *** CCP4 home page http://www.ccp4.ac.uk *** > > > an 11% gap seems worrisome to me. When things are that overfit, I think it > can be harder to find the problems just by looking at how the > model-as-built fits the 2FoFc. There are lots of "bad" ways to get a side > chain into density. You should check the geometry carefully, and if a > residue has poor a fit in the Ramachandran plot, or if it has sub-optimal > side chain dihedrals, first lego the backbone, then lego the side chain, > and you'll probably find a better way to stuff the same thing into the > density. > It seems to me that the more automatic PI's think map-fitting should be, > the more eclipsed leucines land in the databank ... > -Phoebe > > At 09:55 AM 7/5/2006, you wrote: >> *** For details on how to be removed from this list visit the *** >> *** CCP4 home page http://www.ccp4.ac.uk *** >> >> >> Did you refine the model after the molecular replacement? If so, do not do >> so. Look at the maps directly and try to correct for any errors you see, >> especially considering the difference maps (at 3 sigma). >> >> If you run refinement without looking at the maps it would smoothen the >> differences between the MR Model and your data and that is not desired at >> this stage. >> >> Tim >> >> -- >> Tim Gruene >> Institut fuer anorganische Chemie >> Tammannstr. 4 >> D-37077 Goettingen >> >> GPG Key ID = A46BEE1A >> >> >> On Wed, 5 Jul 2006, [utf-8] Li Sheng wrote: >> >>> *** For details on how to be removed from this list visit the *** >>> *** CCP4 home page http://www.ccp4.ac.uk *** >>> >>> >>> Dear Tim Gruene, >>> >>> This structure is a mutant of a protein previously solved. The space >> groups for the wild type protein is P2(1)2(1)2(1) for one crystal and >> C222(1) for the other. The conditions of crystallization for the two >> proteins are different. Of course, this structure is solved by molecular >> replacement. We have not added water molecules now. >>> We have tried both Rafmac and CNS, and the restraints was set to 0.05. >> But we'll try again. Anyway, thank you all. >>> >>> >>> ================= 2006-7-5 15:51, your message: Re: >> [ccp4bb]:================= >>> >>> -----BEGIN PGP SIGNED MESSAGE----- >>> Hash: SHA1 >>> >>> How did you create to this model? Was the gap between R and Rfree this >>> high from the very beginning? If not, try to find out where the gap became >>> so high. >>> >>> - - did you introduce too many water molecules? You can overfit your data >>> this way by pretending noise were waters >>> - - do you have to tighten the weights of stereochemical restraints? If you >>> use refmac5, try setting it to 'AUTO' or something below 0.1. A good >>> indicator off the weight being too high is the list of rmsd's at the end >>> out the refmac5 logfile. >>> >>> Look at your difference map (fofc) at +/- 3 sigma, not 2 sigma. You want >>> that map to show as little density as possible (either 3 or 2 sigma, but 3 >>> is more realistic, I dare say). >>> >>> Tim >>> >>> - -- >>> Tim Gruene >>> Institut fuer anorganische Chemie >>> Tammannstr. 4 >>> D-37077 Goettingen >>> >>> GPG Key ID = A46BEE1A >>> >>> >>> On Wed, 5 Jul 2006, [gb2312] Li Sheng wrote: >>> >>>> *** For details on how to be removed from this list visit the *** >>>> *** CCP4 home page http://www.ccp4.ac.uk *** >>>> >>>> >>>> Hi, dear all, >>>> >>>> Please do me a favour. I ever collected a data set of 2.7A. The space >>>> group is C2. The model fit the 2fofc map and the omit map very well. The >>>> R factor >>>> is 0.26 now but R free is 0.37 and can not be minished. What kind of >>>> error can cause such a gap between R factor and R free? >>>> Thanks in advance. >>>> >>>> >>>> Sincerely yours, >>>> Li >>>> 2006.06.07 >>>> >>> -----BEGIN PGP SIGNATURE----- >>> Version: GnuPG v1.4.1 (GNU/Linux) >>> >>> iD8DBQFEq8PcUxlJ7aRr7hoRAmpBAJ49RvoOuaXYNtD0S/ErcqCctbsHUgCg9+rG >>> 8xzBUsvicmQqcAYD62OU+K4= >>> =Pb/D >>> -----END PGP SIGNATURE----- >>> >>> >>> . >>> >>> >>> >> ============================================================================= >> =============== >>> >>> >>> Sincerely yours, >>> Li Sheng >>> 2006-07-05 >>> __________________________________________ >>> Email:[EMAIL PROTECTED] >>> Institute of Biochemistry and Cell Biology >>> Institutes for Biological Sciences >>> Chinese Academy of Sciences >>> 320 Yue-Yang Road, Shanghai 200031, China >>> Tel: +86-21-5492-1217 >>> __________________________________________ >>> > > ------------------------------------------------------------------------------ > --------------------------------------------- > Phoebe A. Rice > Assoc. Prof., Dept. of Biochemistry & Molecular Biology > The University of Chicago > phone 773 834 1723 > fax 773 702 0439 > http://bmb.bsd.uchicago.edu/index.html > http://www.nasa.gov/mission_pages/cassini/multimedia/pia06064.html >
