If you have any weak anomalous scatterers, even sulfurs, you can use these to improve the phases (use partial model phases in Phaser to find sites, make map based on these, then iterate with autobuilding). This worked well for me recently in a twinned case. With twinning, you have to be very careful about the spacegroup as well, as Tom mentioned. Best option: get better data/crystals, either untwinned or high multiplicity. Read up on twinning as well! JPK From: [email protected] [mailto:[email protected]] On Behalf Of Terwilliger, Thomas Charles Sent: Monday, March 23, 2015 11:54 AM To: Ni, Fengyun; [email protected] Subject: Re: [phenixbb] suggestions on poly-glycine model Hi Fengyun, First it would be good to check very carefully about twinning and space group. What are the values of the statistics listed at the end of Xtriage (Statistics independent of twin laws and Multivariate Z score L-test:)? Are you positive about the space group? Then what is the solvent content for 2 or 4 copies? Yes, it is entirely possible for an incorrect solution to come from mr_rosetta. It will report the best solution it can find, but that might not be correct. I would look very carefully at a composite omit map and an m2Fo-DFc map for your structure, with and without twin laws. If you can see clear things to change, then change them. If it is all very unclear, then you may be stuck for now. All the best, Tom T ________________________________ From: [email protected]mailto:[email protected] [[email protected]] on behalf of Ni, Fengyun [[email protected]] Sent: Sunday, March 22, 2015 3:33 PM To: [email protected]mailto:[email protected] Subject: [phenixbb] suggestions on poly-glycine model Hi everyone, I am working on a ~2.8 A data, Mathew coefficients suggests there are three copies of monomers (194aa) in au. A MR solution (~12% identity with the template) was found by phenix.mr_rosetta in P64 space group. This MR solution contains two copies of model (only 80aa each). But further autobuild cycles on this solution only gives R~0.48. As phenix.xtriage suggests, I add the twin law in the refinement, the R becomes~0.42 but still not succeed in the following autobuild cycles. Then I trim the MR solution to poly-glycine, and do another MR search, and could find two copies of the previous MR solution. Now, the new MR solution contains four copies of search model (80aa each model). With the twin-law, this new MR solution gives R~0.37. Now I am stuck in this step. The autobuild with rebuild_in_place=False or True could not further improve the model or R factors. Any suggestions on the following model building? Or maybe the MR solution I get from phenix.mr_rosetta still not represent the correct one? Thanks in advance! Fengyun