Fri Jun 6 09:16:31 PDT 2008 T0437 Make started Fri Jun 6 09:17:00 PDT 2008 Running on cheep.cse.ucsc.edu Fri Jun 6 12:53:16 PDT 2008 Kevin Karplus This seems to be a homology model for 2j5zA, though there are other templates that seem to match the structure fairly well. Our try1-opt3 model looks ok, except for the final helix (not from an alignment), which seems to be rotated with the buried face out. Fri Jun 6 13:50:01 PDT 2008 Kevin Karplus The rr predictions also suggest that I may have an off-by-2 error in aligning the third strand (an edge strand), since rr constraints pair up a number of residues: I67 L76 K66 D77 Y65 A78 A64 A79 V63 F80 SheetConstraint V63 I67 F80 L76 Hbond A79 while try1-opt3 has SheetConstraint V63 K68 F82 D77 hbond A64 It looks like there are some ssbonds also: C26-C84 C8-C13 For try2, I'll put in the ssbonds, the strand change, and up the burial weights to try to fix these problems. Fri Jun 6 17:15:33 PDT 2008 Kevin Karplus try2 manages to form one of the ssbonds (C8-C13) but doesn't really manage the shifted strand. For try3, I'll do the same optimization, but without shifting the strand. Tue Jun 10 05:42:23 PDT 2008 Kevin Karplus try2-opt3 scores better than try3-opt3, with the try3 costfcn. The big difference is in the breaks. Tue Jun 10 06:03:54 PDT 2008 Kevin Karplus For try4, I've extended the sheetconstraint for the strand that aligned best in try1 from SheetConstraint (T0437)T19 (T0437)D22 (T0437)F30 (T0437)T27 hbond (T0437)V20 1 to ConstraintSet extended_strand SheetConstraint D16 D22 F33 T27 hbond V20 1 I've also turned down the weight on known_ssbonds to keep them from dominating the optimization. I may have to turn that back up again. Tue Jun 10 13:38:34 PDT 2008 Kevin Karplus try4-opt3.gromacs0 scores best with the try4 costfcn. It looks fairly good, except that the helix at Q85-A99 seems to be rotated the wrong way (burying the hydrophilic face). The MQAU evaluation likes the Zhang-Server and the SAM-T08-server models. The MQAC evaluation likes MULTICOM-CLUSTER_TS2 and MULTCIOM-RANK_TS5 models best. I've started metaserver runs with the try1 costfcn. The only reasonable template (2jz5A) is an NMR model that is surprisingly open. Thu Jun 12 11:18:47 PDT 2008 Kevin Karplus MQAC1-opt3 is based mainly on MULTICOM-RANK_TS5 MQAU1-opt3 is based mainly on SAM-T08-server_TS1 Thu Jun 12 15:17:11 PDT 2008 Kevin Karplus I superimposed try1 through try4, MQAC1, MQAU1, and the first 3 alignments. The MQAC1 and MQAU1 models, and the alignment to 2jz5A pretty much agree from D16 on, though the loop S55-V63 varies. I'm not really that fond of these models, though, since A14-I21 is predicted to be a strand, and the try* models do attempt to make a strand out of it. The rr predictions strongly support the idea of the extra strand. Thu Jun 19 12:00:18 PDT 2008 SAM-T08-MQAO hand QA T0437 Submitted Thu Jun 19 12:00:18 PDT 2008 SAM-T08-MQAU hand QA T0437 Submitted Thu Jun 19 12:00:18 PDT 2008 SAM-T08-MQAC hand QA T0437 Submitted Mon Jun 23 14:25:00 PDT 2008 Kevin Karplus The big decision to make on this target is whether C26-C84 and C8-C13 really are disulfide bonds, and if they are, how to get them to form without bad gaps in the backbone. Tue Jun 24 12:16:41 PDT 2008 Kevin Karplus try5 will be a polishing run under the assumpton that the try4 secondary structure is right and that the disulfides are real. Tue Jun 24 13:13:37 PDT 2008 Kevin Karplus try5 polishes try4-opt3, reducing breaks and improving the ssbonds. try5-opt3.gromacs0.repack-nonPC is rosetta's favorite of the repacked models. Tue Jun 24 13:24:09 PDT 2008 Kevin Karplus The models I have seem to fall into a few major groups try5 < try4 < alignments (1j2rA?) has clean hairpin around D22-T27 MQAC1 < MULTICOM-RANK_TS5 N-terminus not part of sheet MQAU1 < SAM-T08-server_TS1 N-terminus not part of sheet try1,try2,try3 slight variants on N-terminus of try4/try5 Maybe I should do metaserver runs with the try5 costfn, to see if there is another model in that family. None of the models quite forms a disulfide for C26-C84, though some manage to form C8-C13. I might want to crank the known_ssbonds up further to force that. I'll wait until the MQAU5 and MQAC5 runs are done before running try6 to try to force SSbonds. Tue Jun 24 14:52:37 PDT 2008 Kevin Karplus MQAU5, mainly from Zhang-Server_TS5 has an interesting different alignment for S55-Q73, that makes a better strand. MQAC5 from MULTICOM-CLUSTER_TS3 is less exciting. Both do not attach the N-terminal strand. I think I'll make a try5-MQAU5 chimera (taking R37-E75 from MQAU5). and optimize that for try6, rather than polishing from all models. Tue Jun 24 15:47:20 PDT 2008 Kevin Karplus try6-opt3 looks better than previous models. I wonder if I can extend the edge strands to tighten things up more. Tue Jun 24 16:30:59 PDT 2008 Kevin Karplus try7-opt3 scores better than previous models, but rosetta thinks that V74, L70, C26, and Q85 all have terrible clashes. undertaker sees these clashes as bad, but not as bad as rosetta thinks they are. I'll do a polishing run with soft_clashes turned up. Tue Jun 24 17:17:56 PDT 2008 Kevin Karplus try8 is better, but still has clashes (breaks are closed, though). Tue Jun 24 17:56:49 PDT 2008 Kevin Karplus try9 reduces clashes a little, opening small breaks, and improving the extended sheets a little. The differences are not visible in a superimposed model, though, so there is no point to further polishing in this direction My major groups are now try9 < try8 < try7 < try6 < chimera-try5-MQAU5 try5 < try4 < alignments (1j2rA?) has clean hairpin around D22-T27 MQAU5 < Zhang-Server_TS5 N-terminus not part of sheet MQAC5 < MULTICOM-CLUSTER_TS3 N-terminus not part of sheet MQAC1 < MULTICOM-RANK_TS5 N-terminus not part of sheet MQAU1 < SAM-T08-server_TS1 N-terminus not part of sheet try1,try2,try3 slight variants on N-terminus of try4/try5 I'll definitely submit try9 and try5, but which 3 other models should I pick? Wed Jun 25 17:18:04 PDT 2008 Firas & John We looked at best-models and liked T0437.MQAU1-opt3.gromacs0.pdb because it is the only one that has the N-terminus packing against the sheet (specifically the hydrophobic resiudes 76 & 78) although it would be better if the helix 17-20 was rotated to bury the hydrophobics better (but it probably doesn't actually matter) It's also the only one that goes in that general region of the protein. Wed Jun 25 17:24:40 PDT 2008 Chirag, Firas & John we all agree: T0437.MQAC5-opt3.pdb is crap, we don't like how the N-terminal end packs horribly when looking with the 'near' script. Residues 18-25 don't seem right at all. we like T0437.try1-opt3.repack-nonPC.pdb because it correctly forms a sheet using strandT0437.MQAU5-opt3.pdb 18-21 (which the last two models we mentioned did not do) The ehl2 script has the strand prediction starting at 13 (instead of 18 according to str2) so we wouldn't mind it if that strand was longer, but overall that region is ok. In T0437.MQAU5-opt3.pdb, a segment (16-23) doesn't pack well with the rest of the structure--there is too much space separating the strand from the rest of the protein and two predicted burried residues (nb11) are exposed. We think this model is garbage. In T0437.MQAC1-opt3.gromacs0.pdb, does not have a strand where str2 predicts one starting at 13. Also, two nb11-predicted burried residues at 15 & 17 are exposed. Thu Jun 26 11:13:32 PDT 2008 Kevin Karplus Summarizing the observations, it looks the top 4 models are ReadConformPDB T0437.try9-opt3.pdb ReadConformPDB T0437.try5-opt3.pdb ReadConformPDB T0437.try1-opt3.repack-nonPC.pdb ReadConformPDB T0437.MQAU1-opt3.gromacs0.pdb # choose one of the remaining models? ReadConformPDB T0437.MQAU5-opt3.pdb ReadConformPDB T0437.try2-opt3.gromacs0.pdb ReadConformPDB T0437.MQAC1-opt3.gromacs0.pdb ReadConformPDB T0437.try3-opt3.gromacs0.repack-nonPC.pdb order with different costfcns: (ignoring try4 and newer) costfcn try1 try1 MQAC1 MQAU5 MQAU1 try2 MQAC5 try3 try2 MQAU5 try2 MQAC5 try3 MQAC1 try1 MQAU1 try3 MQAU5 MQAC5 try2 try3 MQAC1 try1 MQAU1 try4 MQAU5 try1 MQAC5 try2 MQAC1 try3 MQAU1 try5 MQAC5 MQAU5 try2 MQAC1 try1 try3 MQAU1 Sorting just on pred_nb11_back: try9-opt2 try8-opt3 try7-opt1-scwrl try6-opt3 try4-opt3 try5-opt3 try3-opt1-scwrl MQAU5-opt3 try1-opt3 try2-opt3 MQAC5-opt3 MQAU1-opt3 MQAC1-opt3 So the scoring functions pretty consistently don't like MQAU1, but I'll toss it in as model 5 for Firas and John (also, it's from the SAM-T08-server, so is legitimately one of ours). I don't want to use MQAU5, despite its good scores, because the good parts of it were copied and included in try9. So I'll include versions of try1 and try2. These may be a bit too similar to each other and to try9+try5, but I don't really like the ugly meta-server models. 1 T0437.try9-opt3.pdb < try8-opt3 < try7-opt3 < try6-opt3 < chimera-try5-MQAU5 chimera-try5-MQAU5 mainly from T0437.try5-opt3.pdb R37-E75 from MQAU5-opt3 MQAU5 < Zhang-Server_TS5 try5 < try4-opt3 < align (1j2rA) 2 T0437.try5-opt3.pdb < try4-opt3 < align (1j2rA) 3 T0437.try1-opt3.repack-nonPC.pdb < align(2jz5A) 4 T0437.try2-opt3.gromacs0.pdb < align(2gatA) 5 T0437.MQAU1-opt3.gromacs0.pdb < SAM-T08-server_TS1