Hi Sam As I said, if you have NCS (doesn't matter whether you restrained it or not) you need to be aware that the results may not be accurate, because there are well-documented ways in which NCS can affect Rfree completely unpredictably.
With that in mind, if you refer to our Acta D paper http://journals.iucr.org/d/issues/1998/04/00/ad0030/ad0030.pdf and refer to equation 16: y = sqrt((1+ax)/(1-ax)) Here x is the ratio Natoms/Nrefls, y is the expected ratio <Rfree/Rwork> and a=2 for the case of restrained refinement with 4 variables per atom (x,y,z,B) (see text for explanation), so x = (6100+200)/32585 = 0.19334 and therefore y = sqrt((1+2*0.19334)/(1-2*0.19334)) = 1.5036, so since Rwork = 0.182 we get <Rfree> = 1.5036*0.182 = 0.274 with an error range at the 3 sigma level as I said of +/- 0.014. So I would say your actual Rfree = 0.267 is pretty well spot-on. Fig 1 in the paper shows the results for other PDB entries so you can compare. Note however that this by itself doesn't prove that your structure is optimal, all we've done is demonstrate that Rfree is (probably) optimal *assuming that the number of parameters that you used is optimal*, i.e. we took your estimate of 6300 atoms at face value. What happens is, assuming always that the refinement has converged (if it hasn't then it's not possible to draw any conclusions from Rfree), Rfree is sensitive (i.e. increases above its optimal minimum value) to both underfitting (either too few parameters, or the wrong choice of parameters, or refinement stuck in a false minimum) and overfitting (too many parameters) with the ideal at the minimum produced by these competing effects. The above is only a test for underfitting - getting the expected Rfree doesn't rule out overfitting (caused for example by being over-enthusiastic with water addition!) simply because it makes the assumption that the number of parameters you used is optimal, hence you must rule that out by verifying that you can't get a lower Rfree with a different parameterisation, say with fewer waters, then repeat the above test. Hope this helps. -- Ian > -----Original Message----- > From: U Sam [mailto:[EMAIL PROTECTED] > Sent: 22 June 2008 04:47 > To: Ian Tickle > Subject: RE: [ccp4bb] How many reflections for Rfree? > > > Hi Ian, > I have nearly 6100 protein non-hydrogen atoms, 200 waters. > I did not use NCS during the refinement although there are > two molecule in the asymmetric unit. I use C2 space group. > reflections for Rfree are selected randomly in CCP4. > Now I believe you can suggest me something how to proceed. > Thanks. Sam > > > ---------------------------------------- > > Date: Sat, 21 Jun 2008 14:42:49 +0100 > > From: [EMAIL PROTECTED] > > Subject: Re: [ccp4bb] How many reflections for Rfree? > > To: [email protected] > > > > Hi U > > > > Well if you can tell me the total number of atoms in your > PDB file (NOT > > counting any H atoms), I can estimate a range for the > expected optimal > > Rfree assuming your value of Rwork and the no of reflns in > your working > > set (~= 19x1715 = 32585 right?). If you have any NCS I > can't promise my > > estimate will be very accurate, because it will also depend > a lot on the > > nature of the NCS, how you selected your test set, how you > restrained > > the NCS etc. I can already tell you that the range of optimal Rfree > > (+/- 3 sd's) will be ~ +/- 0.014 (i.e. +/- > 3x0.267/sqrt(2x1715)) from > > whatever is the expected value. > > > > Cheers > > > > -- Ian > > > >> -----Original Message----- > >> From: [EMAIL PROTECTED] > >> [mailto:[EMAIL PROTECTED] On Behalf Of U Sam > >> Sent: 20 June 2008 21:23 > >> To: Mark J. van Raaij; [email protected] > >> Subject: RE: [ccp4bb] How many reflections for Rfree? > >> > >> I use CCP4i, refmac5 for the refinement using data of 2.45 > >> angstrom. My R and Rfree is 0.182 and 0.267 respectively. For > >> calculating Rfree ,5% of random data (1715 reflections) was > >> used . So I see there is a difference of about 8.5% between R > >> and Rfree. Is this difference reasonable ? > >> Any idea how can I improve Rfree and difference between R and > >> Rfree gets less than 5%. > >> Thanks, > >> > >> _________________________________________________________________ > >> Need to know now? Get instant answers with Windows Live Messenger. > >> http://www.windowslive.com/messenger/connect_your_way.html?oci > >> d=TXT_TAGLM_WL_Refresh_messenger_062008 > >> > >> > > > > > > _________________________________________________________________ > The i'm Talkathon starts 6/24/08. For now, give amongst yourselves. > http://www.imtalkathon.com?source=TXT_EML_WLH_LearnMore_GiveAmongst > > Disclaimer This communication is confidential and may contain privileged information intended solely for the named addressee(s). It may not be used or disclosed except for the purpose for which it has been sent. If you are not the intended recipient you must not review, use, disclose, copy, distribute or take any action in reliance upon it. If you have received this communication in error, please notify Astex Therapeutics Ltd by emailing [EMAIL PROTECTED] and destroy all copies of the message and any attached documents. Astex Therapeutics Ltd monitors, controls and protects all its messaging traffic in compliance with its corporate email policy. The Company accepts no liability or responsibility for any onward transmission or use of emails and attachments having left the Astex Therapeutics domain. Unless expressly stated, opinions in this message are those of the individual sender and not of Astex Therapeutics Ltd. The recipient should check this email and any attachments for the presence of computer viruses. Astex Therapeutics Ltd accepts no liability for damage caused by any virus transmitted by this email. E-mail is susceptible to data corruption, interception, unauthorized amendment, and tampering, Astex Therapeutics Ltd only send and receive e-mails on the basis that the Company is not liable for any such alteration or any consequences thereof. Astex Therapeutics Ltd., Registered in England at 436 Cambridge Science Park, Cambridge CB4 0QA under number 3751674
