Mon May 26 10:15:39 PDT 2008 T0417 Make started Mon May 26 10:16:31 PDT 2008 Running on peep.cse.ucsc.edu Mon May 26 10:18:18 PDT 2008 Kevin Karplus Crystallographers say "dimer or tetramer." Weak BLAST hit to 1y9kA (E-value 0.15). Mon May 26 12:05:34 PDT 2008 Kevin Karplus Strong HMM hits to d.108.1.1 (1yreA, 2fckA, 2fsrA, 1mk4A, 1yr0A, ...) 1y9kA is a member of this large family, which has at least 100 members. Mon May 26 14:06:06 PDT 2008 Kevin Karplus try1 seems to be working off of 1tiqA, which is only the 15th highest template in the best-scores.rdb file and 22nd in the t2k best-scores file, so is probably not the closest member of the family. I should probably try building a model using 1y9kA and 1yr0A, which score well with both BLAST and HMMs. Mon May 26 15:21:55 PDT 2008 Kevin Karplus The try1 models and the models from alignments are very similar for R26-L149, but the N-terminus is free modeling, and the C-terminus uses quite a different alignment. The try1 alignment looks more convincing to me than the alignment from 2ge3A, which has an uncloseable gap. The alignment to 3c26A looks more feasible. C62 and C69 come close enough to interact, but they aren't conserved, so a metal-binding site seems unlikely. H64, H65, and H67 are nearby, so I wouldn't rule out metal binding. None of the numerous HIS residues are particularly conserved. Mon May 26 17:07:01 PDT 2008 Kevin Karplus try2-opt3, built mainly from 2ge3A, scores better than try1, though the gromacs-optimized model scores better (smaller clashes and breaks). I could try3 as a polishing run from existing models, but I think I'll do another run like try2, but excluding 2ge3A, to get a next-best template. Mon May 26 18:42:03 PDT 2008 Kevin Karplus This looks like the sort of dimer that shares a single sheet---I should optimize as a dimer. Fri May 30 13:29:36 PDT 2008 SAM-T08-MQAO hand QA T0417 Submitted Fri May 30 13:29:36 PDT 2008 SAM-T08-MQAU hand QA T0417 Submitted Fri May 30 13:29:36 PDT 2008 SAM-T08-MQAC hand QA T0417 Submitted The MQAU function prefers SAM-T08-server_TS2, Zhang-Server_TS5, and Zhang-Server_TS3 The MQAC function prefers the 5 Zhang-Server models, and thinks that the SAM-T08-server models are significantly worse (down in the middle of the pack). I'll start metaserver runs for both using try3.costfcn. Tue Jun 3 15:17:57 PDT 2008 Kevin Karplus The try3 costfcn likes MQAU1-opt3, MQAC1-opt3, then try2-opt2.gromacs0 I really don't like the C-terminus of MQAU1-opt3, as it has lost a strand from the sheet. This one comes mainly from SAM-T08-server_TS1. MQAC1 comes mainly from MUSTER_TS5. I'll do a polishing run, excluding MQAU1, to get the rest to have as tight a dry packing. I'll increase breaks and clashes a bit for try4, and put in sheet and helix constraints from try2-opt3 (for Y27 on). Note: the change in weights made try4.costfcn prefer MQAC1-opt3. Tue Jun 3 15:36:31 PDT 2008 Kevin Karplus running try4. Tue Jun 3 17:28:55 PDT 2008 Kevin Karplus The best models from each run with the try4 costfcn are try4-opt3.gromacs0 < MQAC1-opt3 MQAC1-opt3 < MUSTER_TS5 try2-opt3.gromacs0 < 2ge3A MQAU1-opt3 0' failed. The others completed. All four of the costfcns agree that try12-opt3 is the best of these models. The lowest costs come from the try11 cost function, which had the try1 sheets and helices. Sun Jun 8 13:04:25 PDT 2008 Kevin Karplus I've started dimer/try13..try16 to do the optimizations that I had expected of try9..try12, with each starting from only one dimer model---the one its costfcn has the sheet and helix constraints for. Although I'm doing this polishing in order to get multiple models, the try12-opt3 model actually does look pretty good to me. Sun Jun 8 13:11:30 PDT 2008 Kevin Karplus OOPS. Bad bug: dimer/costfcn-init.under had KnownBreak 190 instead of KnownBreak S190. This has made all the break costs depend heavily on the inter-monomer break! Sun Jun 8 13:24:50 PDT 2008 Kevin Karplus Rescoring the dimers indicates that all of try9-try12 costfcns prefer try7-opt3. In other words, the flaw in the costfcn was definitely emphasizing the wrong thing. I'll also do a dimer/try17 run to reoptimize the dimer from try5. Sun Jun 8 15:56:53 PDT 2008 Kevin Karplus costfcn Favorite dimer models dimer/try17 try7-opt3 try13-opt3 try8-opt3 try17-opt3 try11-opt3 dimer/try16 try13-opt3 try7-opt3 try8-opt3 try17-opt3 try11-opt3 dimer/try15 try13-opt3 try7-opt3 try8-opt3 try17-opt3 try11-opt3 dimer/try14 try13-opt3 try7-opt3 try8-opt3 try17-opt3 try11-opt3 dimer/try13 try13-opt3 try7-opt3 try8-opt3 try17-opt3 try11-opt3 That seems like prety close agreement on which models are most reasonable! Note that try7-opt3 does not form the desired dimer interface, so I should do a resoding that increases the weight of the dimer interface. I made a try18 costfcn that has constraints for the try13 sheets and helices, but ups the dimer, clashes, and breaks. costfcn Favorite dimer models dimer/try18 try13-opt3 try8-opt3 try17-opt3 try11-opt3 try9-opt2 try13-opt3 < dimer-try6-2ge3_1 try8-opt3 < dimer-try5-2ge3_1 try17-opt3 < dimer-try5-2ge3_1 try11-opt3 < dimer-try5-2ge3_1 try9-opt2 < dimer-try5-2ge3_1 So these really only represent two of the initial dimers, and even with the try13 (from try6) sheets and helices, the try5 models do well. For dimer/try18 and dimer/try19, I'll up the weight of the sheet and helix constraints (based on try13 and try8 respectively) and see if I can produce a more polished model. Sun Jun 8 16:22:24 PDT 2008 Kevin Karplus After modifying the try18 costfcn to really crank up the weights on the sheet and helix constraints: costfcn Favorite dimer models dimer/try18 try13-opt3 try17-opt3 try8-opt3 try11-opt3 try7-opt3 try9-opt2 dimer/try19 try17-opt3 try8-opt3 try11-opt3 try7-opt3 try9-opt2 try12-opt3 I don't think I'll bother polishing up the less-good models: try14 < dimer-MQAU1-2ge3_1 try15 < dimer-try1-2ge3_1 try16 < dimer-try3-2ge3_1 I think they roughly rank in that order. Sun Jun 8 18:36:09 PDT 2008 Kevin Karplus costfcn Favorite dimer models dimer/try18 try18-opt3 try13-opt3 try17-opt3 try8-opt3 try11-opt3 try7-opt3 try9-opt2 dimer/try19 try19-opt3 try17-opt3 try8-opt3 try11-opt3 try7-opt3 try9-opt2 try12-opt3 I could try squeezing out the breaks a bit more, but I doubt that I'll really make any significant improvement at this point. I could also try building dimers off of other templates (such as 1yr0A), but I'm not sure it is worth the effort. I'll submit InFilePrefix dimer/decoys/ ReadConformPDB T0417.try18-opt3.unpack.pdb chain A ReadConformPDB T0417.try19-opt3.unpack.pdb chain A ReadConformPDB T0417.try14-opt3.unpack.pdb chain A ReadConformPDB T0417.try15-opt3.unpack.pdb chain A ReadConformPDB T0417.try16-opt3.unpack.pdb chain A