Aug 16 17:24:22 PDT 2004 T0279 Mon Aug 16 20:29:18 PDT 2004 Kevin Karplus Comparative model with 1jr2A as template. I've set up MANUAL_TOP_HITS with the best template and the top few next-best hits from both t2k and t04 searches. Tue Aug 17 12:53:48 PDT 2004 Kevin Karplus try1-opt2 looks pretty good, except that the two domains don't pack against each other, despite a solid buried patch on each. That lack of packing is coming directly from the template, so I don't really want to mess with it. Maybe, I'll just do some polishing with an unconstrained cost function and submit. Try2 is running to do this polishing. There are strong RR constaints across the domain boundary, though, so it might be worth turning them on and seeing what happens to the domains. Perhaps we could have both an open and a closed conformation? Undertaker is not too good at hingeing motions, so this may be difficult. Tue Aug 17 17:37:19 PDT 2004 Kevin Karplus The try2-opt2 model looks fairly good, but the attempt to close in the try3 model was unsuccessful, tearing the sheets, rather than folding the linker. The unconstrained costfcn reflects this judgement, with try2-opt2 scoring well, but try3-opt2 being worse than try1-opt2. try2-opt2 has one huge break at N213-T214. I think I need another polishing run with break weight higher. There is also rather ugly break at E20-A21, though it is only 5th on the list of breaks by cost. I think I'll just work on the open conformation, and not try to fold the domains together. Tue Aug 17 21:49:56 PDT 2004 Kevin Karplus The try4 run scores better, though breaks are still a bit big, with still a huge break before T214. Perhaps I need to add a constraint on the residues around the gap to apply more powerful operators? We could do a N213.C T214.N constraint of 1.329 or a N213.CA T214.CA constraint of 3.803. We may have to relax the helix constraints a bit, to let the helix unwind to make the connection. Hmm, there were no constraints in try4, so that means that only the alpha values and hbond_geom_backbone were holding N213 in the helix. I'll add constraints like this (C-N and CA-CA) for the worst 4 breaks in try4 and do a polishing run. Wed Aug 18 12:38:18 PDT 2004 Kevin Karplus The try5 run did close the breaks properly, and now scores best with the unconstrained cost fcn. Rosetta now likes try5-opt2.repack-nonPC best. I could submit now, or I could do one polishing run. The differences between our top models are pretty small, and further gap closure is unlikely to produce really better models, so I'll stop here. I'll submit try5-opt2 best unconstrained try5-opt2.repack-nonPC best Rosetta energy try1-opt2 full auto T0279-1jr2A-t04-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 T0279-1pea-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 Thu Nov 18 20:07:23 PST 2004 Martina Koeva Here are the smooth GDT summary evaluations: best sam-t04 33.0407 best submit 32.6751 model1 32.6751 auto 32.6669 align 31.7320 robetta best 34.5969 robetta1 33.1210 Robetta beat us on this one by just a little, but most of the models are within very close range, except for our align2 (and model 5), which has a score that is quite a bit lower: smooth_gdt = 12.0252