Thu May 29 09:07:37 PDT 2008 T0423 Make started Thu May 29 09:35:34 PDT 2008 Running on moai04.kilokluster.ucsc.edu Make started Thu May 29 09:26:29 PDT 2008 Running on moai08.kilokluster.ucsc.edu Thu May 29 09:22:52 PDT 2008 Kevin Karplus Homology model, 35% identity for almost full length to 2otmA (blast has E-value 1e-18). Thu May 29 10:00:38 PDT 2008 Kevin Karplus This looks indeed like d.79.1.1 proteins, with 2otmA as the best template, but about 10 others available in the t06 template library. Looking just at the t06.w0.5 run, 2dyyA comes out as the top hit, with about 19 templates all told. 2otmA is not top scoring with just the w0.5 models. I may want to add the top hits with the w0.5 searches to MANUAL_TOP_HITS, to get more alignments to model from. Sun Jun 8 11:06:28 PDT 2008 SAM-T08-MQAO hand QA T0423 Submitted Sun Jun 8 11:06:28 PDT 2008 SAM-T08-MQAU hand QA T0423 Submitted Sun Jun 8 11:06:28 PDT 2008 SAM-T08-MQAC hand QA T0423 Submitted Sun Jun 8 21:42:54 PDT 2008 Kevin Karplus The big decision is whether residue K10-T15 form a strand as they do in the second and third alignments or not, as in the first alignment. Doesn't have strand T0423-2otmA-t06-local-str2+near-backbone-11-0.8+0.6+0.8-adpstyle5.a2m Has strand T0423-1x25A-t06-local-str2+near-backbone-11-0.8+0.6+0.8-adpstyle5.a2m:1x25A T0423-1jd1A-t06-local-str2+near-backbone-11-0.8+0.6+0.8-adpstyle5.a2m:1jd1A 2otmA is part of a trimer, and one of the other monomers interferes with where the extra strand of the sheet would go. The Metaservers favor MQAC: RAPTOR_TS2 RAPTOR_TS1 HHpred2_TS1 HHpred4_TS1 MQAU: Zhang-Server_TS5 MULTICOM-CLUSTER_TS2 SAM-T08-server_TS1 All of these have the 5-strand sheet, except Zhang-Server_TS5, which has a 6-strand sheet. Mon Jun 9 05:24:17 PDT 2008 Kevin Karplus The MQAU1-opt3 model is based mainly on SAM-T08-server_TS1 The MQAC1-opt3 model is based mainly on BAKER-ROBETTA_TS2 The best-scoring models with try1.costfcn are MQAC1-opt3 try1-opt3 MQAU1-opt3 all of which have the 5-strand sheet. For try2, I think I'll build models excluding 2otmA as a template, and see what comes up, though the alignment to 2otmA *is* very good, and is likely to be correct. Even the P18, which I thought might favor the turn to come back for the next strand is conserved in 2otmA. Mon Jun 9 08:00:32 PDT 2008 Kevin Karplus For the 5-strand version, I should optimize as a trimer---this should be my top model. For the 6-strand version I should optimize as a monomer. I should also do crossover to improve the 6-strand model, as it has worse breaks than the 5-strand models do. I'll have to create a costfcn that prefers the 6-strand model by enough to keep from falling into the (probably global) minimum for the 5-strand model. Mon Jun 9 13:57:09 PDT 2008 Kevin Karplus try3-opt3 is looking fairly good for the 6-strand option, except for bad breaks before I90 and R91. One more polishing run with try3 helices and strands and high break costs should finish off that option. Then I'll have to set up the trimer for the other option. Mon Jun 9 14:13:37 PDT 2008 Kevin Karplus try4 started to optimize 6-strand model. Still haven't set up trimer for 5-strand model. Mon Jun 9 20:45:13 PDT 2008 Kevin Karplus The trimer/try5-opt3 for the 5-strand model did not move things much. I'm not convinced that the rotamers for F106 and H110 are right in MQAC1-opt3 and dimer/try5-opt3. I'll make another trimer from try1-opt3 and try optimizing that. Mon Jun 9 20:54:14 PDT 2008 Kevin Karplus trimer/mult6 started on moai cluster. I like the trimer/try6 positioning of F106 better. Maybe I need to polish it a bit more, with clashes turned up, to make it score as well as the try5 trimer. try7-opt3.unpack.gromacs0 scores almost as well as try5-opt3.unpack.gromacs0 on the try7 costfcn. I think I'm ready to submit on this model: InfilePrefix trimer/decoys/ ReadConformPDB T0423.try7-opt3.unpack.pdb chain A # favorite 5-strand model