Mon Jul 19 10:26:22 PDT 2004 T0244 Due 14 Aug Mon Jul 19 16:55:24 PDT 2004 Kevin Karplus This looks like a comparative model with 1mp4A (c.68.1.6) as the template There are LOTS of available templates. Tue Jul 20 10:16:53 PDT 2004 Kevin Karplus try1 looks pretty good. The rr constraints are mostly correct. The N- and C-terminal helices need to pack better against the rest. Perhaps we should try an unconstrained polishing run with the burial weights, clashes, and breaks turned up. Tue Jul 20 19:22:26 PDT 2004 Kevin Karplus try2 also looks ok and scores slightly better than try1, but the long helices still haven't packed against the main body, and V185 seems to be waving around in the solvent when it could fold up agains the main body. I wonder if I should just up the dry terms and see what happens, or if I should add specific constraints to try to pin things down. Perhaps V185.CG2-A255.CB? Maybe L293.CG-M154.SD? maybe H282.ND1-Y43.OH? maybe L81.CG-I103-CB? Wed Jul 21 09:01:30 PDT 2004 Kevin Karplus Foo! try3 was just a rerun of try2---I forgot to change the "try2" to "try3" in the script. I renamed the output files in decoys, and restored the missing try2 outputs from my laptop (which happened to have copies). Wed Jul 21 11:01:34 PDT 2004 Kevin Karplus Another goof---not quite so big. The t04.many.frag fragment library had not finished building when try4 tried to read it, so it wasn't included in try4. When try4-opt1 finishes, I should probably kill try4 and start again as try5. Thu Jul 22 15:12:32 PDT 2004 Kevin Karplus try4 ran to completion and scores best with the try4 cost function, but still doesn't pack the helices. The Cterminal K301 seems to get stuck on E92 (though not close enough for a salt bridge), and E92-Q94 interact with K301 and K300. I rather doubt that this interaction is real, and am surprised that it is scoring well enough to interfere with packing the helices. There has been essentially no change since try1. I should probably start over from the alignments, to try to get a version that does not have the helices in this awkward but apparently immovable position. Ahh--one obvious problem: a typo in the try4.costfcn file made dry5 have a weight of 205 instead of 20. This results in this one, not very useful, component dominating the cost function. Fixing the problem in try5.costfcn makes try3 be the best scorer. There are no RR constraints involving either helix, so the eyeball constraints I added before will have to do for guiding the packing. Thu Jul 22 20:47:50 PDT 2004 Kevin Karplus try5-opt2 scores better than try3 with the try5 costfcn, but Rosetta still likes try3-opt2.repack-nonPC better (though it doesn't like any of the models much). Try5 scores best with an unconstrained costfcn also Sat Jul 31 05:11:58 PDT 2004 Kevin Karplus The helices L285-end and S66-E82 are still not packed against the rest very well. I wonder if I should comment out the read-pdb.under and try a run just from the alignments. Mon Aug 2 11:30:56 PDT 2004 Kevin Karplus The try6 score function likes all previous runs better than try6. It looks like this protein has a large hydrophobic pocket that might collapse if we set the dry weights up too high. I wonder if I should add some scaffolding to keep the pocket large? The helices are still not packing against the main body. Tue Aug 3 04:07:43 PDT 2004 Kevin Karplus For try7 I added some scaffolding to try to keep what looks like the active site open and added one constraint to try to improve helix packing. Wed Aug 4 05:45:45 PDT 2004 Kevin Karplus The try7 model is a slight polishing of the try5 model, with no further packing of the helices. Sat Aug 7 14:49:07 PDT 2004 Kevin Karplus I'll make one more attempt at polishing, with stronger constraints to try to get the helices to pack, then a final run without constraints. Sat Aug 7 20:22:43 PDT 2004 Kevin Karplus try8 gets slightly better cost than try7, but it is still not moving the helices. I'll do one last polishing run without constraints, then call it quits---I don't think I'm going to get the helices to move, and it isn't worth further effort. Sun Aug 8 06:46:01 PDT 2004 Kevin Karplus Further optimization in try9 made almost no changes---time to give up and submit try9-opt2 try7-opt2.repack-nonPC try1-opt2 T0244-1mp4A-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 T0244-1lvwA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 Thu Nov 18 23:41:38 PST 2004 Martina Koeva Based on the smooth gdt scores: best sam-t04 52.6974 (this was actually our align2 model, as well model5) best submit 52.6894 (still model5) model1 44.4645 auto 44.1926 align 52.3265 robetta best 50.3346 (robetta model4) robetta1 49.9761