My favorite test for truth in molecular replacement solutions is whether or not the map shows density for something that you know about but the computer doesn't. Take out a base pair or two, or delete some nice fat side chains, and see if they still show up in the density when you use that model to phase the map. If they do, its far more convincing (to me at least) than a page full of statistics. Phoebe
===================================== Phoebe A. Rice Dept. of Biochemistry & Molecular Biology The University of Chicago phone 773 834 1723 http://bmb.bsd.uchicago.edu/Faculty_and_Research/01_Faculty/01_Faculty_Alphabetically.php?faculty_id=123 http://www.rsc.org/shop/books/2008/9780854042722.asp ---- Original message ---- >Date: Mon, 25 Jul 2011 15:38:02 +0800 >From: CCP4 bulletin board <CCP4BB@JISCMAIL.AC.UK> (on behalf of Hubing Lou ><louhub...@gmail.com>) >Subject: Re: [ccp4bb] Phaser and Molrep gave different solutions >To: CCP4BB@JISCMAIL.AC.UK > > Hi, > > Thanks for those who replied to this thread. > I have been trying all the means that people > suggested: search protein alone, DNA alone. However, > both not working out. > One thing Ray Brown suggested "MR works if the > molecules have identical sequences". So I just > played around with the following way: 1) use the > chainsaw editted model in MR; 2) mutate the protein > sequence back to my protein and refine the solution > in Coot; 3) use the partially refined protein-DNA as > the new search model and run Phaser again then I get > the following results for the two copies in ASU: > RFZ=6.5 TFZ=10.7 LLG=903 RFZ=6.2 TFZ=17.4 LLG=1130. > After one round of rigid body and restraint > refinement in Refmac5, the Rfac and Rfree drops from > 0.50/0.51 to 0.46/0.51. I did not refine the DNA > sequences yet. > > I am not sure if this way I can further refine the > structure, or do I bring two much bias into the > structure. Please correct me if any. Thanks. > > Best, > Hubing > > On Thu, Jul 21, 2011 at 11:31 PM, > <ray-br...@att.net> wrote: > > I have also tried for years to solve a protein-DNA > complex without sucess. If you have a lot more DNA > than protein in the AU then MR will not work. > You always get a good RFZ score but you cannot > solve the translation if the DNA molecules are > forming long stacks. With a plausible packing you > will of course get model phases and a nice map but > refinement will not work and you will get stuck at > 40-50% Rf. > > You may have a chancewith MR if you only have a > small DNA and a much bigger protein molecule or if > the search models and the molecules have > identical sequences. > > To solve this structure you probably have to do > Se-labeled protein and SAD etc. or collect > anomalous from metal ions if present. > > Cheers. > > Ray Brown > > ------------------------------------------------ > > From: Hubing Lou <louhub...@gmail.com> > To: CCP4BB@JISCMAIL.AC.UK > Sent: Thu, July 21, 2011 6:39:49 AM > Subject: Re: [ccp4bb] Phaser and Molrep gave > different solutions > I was worried as well with the low TFZ score. > Usually successful cases with score >8. I am still > puzzled why Phaser and Molrep gave different > solutions. Does this mean molecular replacement do > not work out in this case so more crystals have to > be prepared? > > A little more information might be helpful to > dissolve the problem here. The model I used is a > protein-DNA complex. The protein was Chainsaw > editted but the DNA sequence was directly borrowed > from the original model. > > Best, > Hubing > > On Thu, Jul 21, 2011 at 3:30 PM, Vellieux Frederic > <frederic.velli...@ibs.fr> wrote: > > Hi, > > It's not a bad idea to read the Phaser manual > for molecular replacement; see > http://www.phaser.cimr.cam.ac.uk/index.php/Molecular_Replacement > > Soon after the start, in a table on the right > hand side, there is: TFZ score < 5, have I > solved it ? No. > > Hence with a TFZ score of 3.8 you do not have a > solution using Phaser. > > Fred. > > Hubing Lou wrote: > > Dear all, > > I am stuck in a molecular replacement case and > looking for advices. > I have been working on a protein-DNA complex > structure. > Data was processed by HKL2000 to 2.6Ang and > some of the data statistics are shown below: > > Space group: P21, > Unit Cell: 54.73, 104.91, 78.40, 90, 100.2, 90 > Redundancy: 2.8 (2.7) > Completeness: 94.8 (93.1) > Linear R-fac: 0.051 (0.442) > > Data quality was checked by Phenix.xtriage and > there's no problem. I then prepared a model by > Chainsaw. Our protein shares only 30% of > sequence similarity with the model, but > structurally they are in the same group and > almost identical in apo form. Matthrews Coeff > indaced two monomers in AU. I then ran Phaser > in "automated search" mode and there's a > solution with RFZ score 4.8, TFZ score 3.8. > The electron density map was not bad with DNA > double helix clearly seen. However Refmac5 > couldn't get Rfree lower than 50%. > > I then changed to MolRep, ran "self rotation > function" first then used the first 10 peaks > for translation search. Again there's a > solution but it is different from that from > Phaser. I attached a picture here. Checking in > coot, the packing is the same. But, the > refinement couldn't get Rfree lower than 50%. > > I have tried to include NCS, TLS refinement in > Refmac, both not working. > Hope someone out there can help. > Thanks very much for your time. > > Hubing > > > ------------------------------------------------------------------------
