Hi, Those are fairly high LLG values, and the jump when you add the fourth molecule is big. R-factors are very insensitive for unrefined models from templates that don't have high sequence identity, so it's not that surprising that both R-factors are in the random range. I would worry more about the failure of refinement to reduce the R-factors than about how high they are immediately after molecular replacement. You can get big LLG values either by having a correct solution with a reasonably good model, or by reproducing the translations of translational NCS with an otherwise wrong solution. Does xtriage show a big peak in the native Patterson? Assuming you have data to a reasonable resolution, 3 or 4-fold averaging should give you a good map to rebuild the structure, and NCS restraints should help the refinement. But if the resolution is poor or if there's some other issue (e.g. twinning or translational NCS), then it could be more difficult to proceed. You could send me the logfile from the Phaser part of the AutoMR run (preferably off-line), and I could see if there's anything that would worry me. Regards, Randy Read On 27 Aug 2009, at 15:54, Yuan Cheng wrote:
Hi, I am working on solving a protein structure using phenix.automr. There is likely four monomers in a ASU based on phenix.xtriage.I am using a search model with 30% identity with my protein. After finding three monomers in a ASU, the LLG is 1740 and R factor is 56%. After finding the fourth monomer, the LLG increased to 2080,but the R factor also increased to 58%. I have several question here: 1) how much should LLG and R factor be if the model from molecular replacement is correct? I guess the answer might not be a exact number. But I am still wondering what should be the correct range. 2) Whether R factor and LLG should always be positively correlated?
I used phenix.refine to refine the model with 3 monomers and the model with 4 monomers in a ASU. The Rwork and Rfree are similar for these two models,47% and 53% respectively. I think that one more monomer should either improve or ruin the model.I don't quite understand why one extra monomer doesn't matter too much here. In addition, do you think it is worthy to keep doing model building and refinement in consideration such high R factors. All you suggestions and opinions will be greatly appreciated!
Yuan _______________________________________________ phenixbb mailing list [email protected] http://www.phenix-online.org/mailman/listinfo/phenixbb
------ Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: + 44 1223 336500 Wellcome Trust/MRC Building Fax: + 44 1223 336827 Hills Road E-mail: [email protected] Cambridge CB2 0XY, U.K. www- structmed.cimr.cam.ac.uk