Wed May 28 09:04:29 PDT 2008 T0421 Make started Wed May 28 09:05:59 PDT 2008 Running on peep.cse.ucsc.edu Wed May 28 09:07:43 PDT 2008 Kevin Karplus Best BLAST e-value 0.29 for 2bmbA This is supposed to be a polyphosphate kinase2, which means that I expect it to have a similar structure to T0419, even though this protein is substantially shorter. Wed May 28 17:36:20 PDT 2008 Kevin Karplus strong conservation signals, starting with L71 through L274, with particularly strong conservation of P250, W251. The t2k and t06 alignments disagree a bit on which residues are conserved (though not on the PW). Strong agreement on c.37.1.1 domains (as expected from the polyphosphate kinase label), with the more sensitive t04 and t06 searches being more confident than the t2k searches. The try1-opt3 model looks fairly good for G82-K279 (though with somewhat large breaks), matching the conservation on the t2k multiple alignment. We may want to do the M1-G82 region as a separate domain. Wed May 28 17:53:17 PDT 2008 Kevin Karplus M1-G82 prediction started on moai cluster. There is no CYS or HIS clustering, so I should turn off maybe_metal and maybe_ssbond. Sat Jun 7 22:01:16 PDT 2008 Kevin Karplus The MQA evaluations like the Zhang-Server models best, though MQAU likes SAM-T08-server also. I've started the undertaker optimizations for the MQAC and MQAU metaserver runs. Sat Jun 7 22:15:38 PDT 2008 Kevin Karplus I just got around to looking at the M1-G82 predictions. There does not seem to be much in the way of templates (best e-value 37). About all I get out of this prediction is that there is high conservation for Y49 on, and a pretty strong prediction for a helix from R57 to A80. I don't see how the rather poor M1-G82 predictions are going to help much, since we already have the strong helix prediction, and not much else is coming up. The question is where to put the helix. I should do some polishing runs for the full-length undertaker models (without the metaserver models) to see if I can reduce breaks. Sat Jun 7 22:28:22 PDT 2008 Kevin Karplus try2 started on moai cluster to try to close breaks. Sun Jun 8 07:56:38 PDT 2008 Kevin Karplus MQAU1-opt3 is based mainly on SAM-T08-server_TS1 MQAC1-opt3 is based mainly on Zhang-Server_TS4 The try2 costfcn prefers try2-opt3.gromacs0 MQAU1-opt3 MQAC1-opt3 try1-opt3.gromacs0 The main part, roughly V85-D277, is in pretty close agreement among models, though the loop near R210-W230 does differ. There are some pretty terrible breaks in all the models, so I'll try doing another optimization run from all models with breaks turned up. The models are in agreement on the sheet, except for the strand that is about A113-T119, which has different alignments in different models. I'll leave out the sheet constraints for that strand. Sun Jun 8 09:34:39 PDT 2008 Kevin Karplus The hits are to "Nucleotide and nucleoside kinases" c.37.1.1, which are numerous. All the top scores are for c.37.1.* superfamily, at least for the first 90 hits (out to E-value 2). Sun Jun 8 10:58:40 PDT 2008 Kevin Karplus try3-opt3 reduced, but did not eliminate, breaks. It initially worked on MQAC1-opt3 (for try3-opt1 and try3-opt2) but then switched to try2-opt3.gromacs0 (for try3-opt3). Sun Jun 8 11:06:27 PDT 2008 SAM-T08-MQAO hand QA T0421 Submitted Sun Jun 8 11:06:27 PDT 2008 SAM-T08-MQAU hand QA T0421 Submitted Sun Jun 8 11:06:27 PDT 2008 SAM-T08-MQAC hand QA T0421 Submitted Sun Jun 8 11:12:19 PDT 2008 Kevin Karplus Maybe I should do some optimizations with the MQAU1-opt3 sheets and the MQAC1-opt3.sheets also, to see what comes up on top. I'll have to set breaks still higher to try to close gaps. Sun Jun 8 11:24:11 PDT 2008 Kevin Karplus try4 started with MQAU1 sheets and helices try5 started with MQAC1 sheets and helices Both start from a subset of models: ReadConformPDB T0421.MQAC1-opt1-scwrl.pdb ReadConformPDB T0421.MQAC1-opt1.pdb ReadConformPDB T0421.MQAC1-opt2.pdb ReadConformPDB T0421.MQAC1-opt3.pdb ReadConformPDB T0421.MQAU1-opt1-scwrl.pdb ReadConformPDB T0421.MQAU1-opt1.pdb ReadConformPDB T0421.MQAU1-opt2.pdb ReadConformPDB T0421.MQAU1-opt3.pdb ReadConformPDB T0421.try1-init.pdb.gz ReadConformPDB T0421.try1-opt1-scwrl.pdb.gz ReadConformPDB T0421.try1-opt1.pdb.gz ReadConformPDB T0421.try1-opt2.pdb.gz ReadConformPDB T0421.try1-opt3.gromacs0.pdb.gz ReadConformPDB T0421.try1-opt3.gromacs0.repack-nonPC.pdb.gz ReadConformPDB T0421.try1-opt3.pdb.gz ReadConformPDB T0421.try1-opt3.repack-nonPC.pdb.gz ReadConformPDB T0421.try3-opt1-scwrl.pdb.gz ReadConformPDB T0421.try3-opt1.pdb.gz ReadConformPDB T0421.try3-opt2.pdb.gz Sun Jun 8 16:37:44 PDT 2008 Kevin Karplus costfcn preferred models try5 try5-opt3.gromacs0.repack-nonPC try4-opt3.gromacs0.repack-nonPC try3-opt3.repack-nonPC try4 try5-opt3.gromacs0.repack-nonPC try4-opt3.gromacs0.repack-nonPC try3-opt3 try3 try5-opt3.gromacs0.repack-nonPC try3-opt3 try4-opt3.gromacs0.repack-nonPC try2 try5-opt3.gromacs0.repack-nonPC try3-opt3 try4-opt3.gromacs0.repack-nonPC try1 try5-opt3.gromacs0.repack-nonPC try3-opt3 try4-opt3.gromacs0.repack-nonPC rosetta likes try5-opt3.gromacs0.repack-nonPC try3-opt3.gromacs0.repack-nonPC try4-opt3.gromacs0.repack-nonPC try4 worked mainly on MQAU1-opt3 < SAM-T08-server_TS1 try5 worked mainly on try3-opt2 < MQAC1-opt3 < Zhang-Server_TS4 try3-opt3 < try2-opt3.gromacs0 < try1-opt3 I think I may need to do a polishing run, to pick up the changes made by gromacs and rosetta and pack things a little tighter. The big decision to make is about the strand near P110-V119, as the three main models have different phases for it. I don't like the way try4-opt3 crumples it, but that still leaves try3-opt3: SheetConstraint (T0421)F144 (T0421)R149 (T0421)A113 (T0421)L118 hbond (T0421)D148 1 try5-opt3: SheetConstraint (T0421)V115 (T0421)T119 (T0421)F144 (T0421)D148 hbond (T0421)L118 1 which is a simple shift of 2 in strand placement. The try5-opt3 model seems to have a less strained connection to the helix between the strands, so I'll do a polishing run for that alignment. (Maybe I should do one for the try3 alignment also, just to see if it can be cleaned up.) Sun Jun 8 17:22:19 PDT 2008 Kevin Karplus try6 started to clean up try5 (using all existing models) try7 started to clean up try3 (using all models except try5 models) Sun Jun 8 19:26:54 PDT 2008 Kevin Karplus costfcn favorite models try7 try6-opt3 try5-opt3.gromacs0.repack-nonPC try7-opt3 try3-opt3 try4-opt3.gromacs0.repack-nonPC try6 try6-opt3 try5-opt3.gromacs0.repack-nonPC try7-opt3 try4-opt3.gromacs0.repack-nonPC try3-opt3 I have basically three models, differing mainly in the N-terminal residues: try6-opt3 try7-opt3 try3/opt3/try4-opt3 I'm wondering if I should make another attempt at folding the N-terminal residues independently, to try to get something more compact. I could also look at all 90 of the c.37.1.1 templates to see if anything else has that initial helix. I'll do a quick check with VAST, though that will only find helices with the same orientation as in the query structure (in this case try6-opt3). Sun Jun 8 20:00:32 PDT 2008 Kevin Karplus Nope. none of the VAST hits extend back before T81. The top VAST hit is for 2ccjA, which was our 4th highest hit. Other high hits were high on our list also: template vast our rank rank 2ccjA 1 4 4tmkA 2 6 2pbrA 3 1 2z0hA 4 9 1nmyA 5 --- 2plrA 6 10 1g3uA 7 --- 2qg6A 8 --- 2zi3A 9 --- Some of the differences are just in what representatives were chosen for the low-redundancy set. Thu Jun 12 10:24:26 PDT 2008 Kevin Karplus I ran get-seqs -ids T0421.best-scores.rdb -seq stride-mixed.seq T0421.best-scores.seqs -seq stride-mixed.str2 T0421.best-scores.str2s to get the sequences and str2 sequences for our top hits, to see if any have the initial helix. (get-seqs is in /projects/compbio/experiments/models.97/scripts) Maybe 1gvnB, 1m7gA, 1rz3A ? Thu Jun 12 10:40:46 PDT 2008 Kevin Karplus Perhaps I should try adding a constraint to make L67 close to L118 and Q75 close to V116, to try to pack the helix into the somewhat hydrophobic pocket available for it. I'd have to turn clashes and breaks way down to allow movement. 1gvnB does have that helix packing, while 1rz3A puts the initial helix on the other side of the sheet, in a place that is already occupied in our target. For try8, I'll add 1gvnB alignments to the mix, and make a model from alignments. I'll keep the sheet and helix constraints from before, but add distance constraints for the packing I want to see. Thu Jun 12 15:34:57 PDT 2008 Kevin Karplus try8 failed miserably, because it never read in the 1gvnB alignments. I'll try again as try9 and try10 (same costfcn, but try9 will have 1gvnB in addition to the other templates, and try10 will have only 1gvnB). Thu Jun 12 19:33:26 PDT 2008 Kevin Karplus try10 put the big helix roughly where I told it to, but only by leaving an unclosable gap between K83 and R84. If I unwound a turn of the helix, back to Q79 it might just fit. Even so, the angle of the helix is not what I would want. I'm still trying to figure out whether time spent packing this helix is worthwhile. Sun Jun 15 09:13:00 PDT 2008 Kevin Karplus I'll try doing a try11 run from all models, to see if I can get better packing with clashes and breaks turned way down, and with sheet constraints from both try4 and try6. Since I don't have any constraints attempting to improve the packing, I suspect that I'll just tweak up try6 a bit, maybe making it better, maybe worse. Sun Jun 15 10:25:18 PDT 2008 Kevin Karplus try11-opt1 and try11-opt2 made some sort of improvement to try6-opt3 while increasing clashes and breaks, but try11-opt3 ignored those and went back to working on try6-opt3, decreasing breaks slightly. When try11 finishes, I'd like to do another run using the new DoubleCross operator, to see if it ever helps. Sun Jun 15 11:54:54 PDT 2008 Kevin Karplus try11-opt3 increased clashes and breaks slightly over try6-opt3, but improved Hbonds and burial measures. Rosetta does not like it as much (probably because of the slightly greater clashes). I'd really like to pack in the N-terminal helices on one side or the other of the P166-A184 helix, but I've not had much luck doing this with undertaker. This is where I could use a tool like FoldIt! Tue Jun 17 20:40:24 PDT 2008 Kevin Karplus I've been busy with other targets and not had time to come back to T0421. Currently, I'm looking at submitting ReadConformPDB T0421.try6-opt3.pdb < try5.gromacs0.repack-nonPC