Tue Jun 13 09:38:29 PDT 2006 T0328 Make started Tue Jun 13 09:39:30 PDT 2006 Running on lopez.cse.ucsc.edu Tue Jun 13 09:40:44 PDT 2006 Kevin Karplus Excellent blast hit to 2gvkA (34% over 306 residues, 1.8e-46) Make started Tue Jun 13 18:14:05 PDT 2006 Running on lopez.cse.ucsc.edu Restarted the job now that oldmacd is fixed and /projects/compbio/tmp is available again. Make started Tue Jun 13 21:20:53 PDT 2006 Running on lopez.cse.ucsc.edu Tue Jun 13 21:23:48 PDT 2006 Kevin Karplus Found a minor bug in Make.main, so restarted the make. Sat Jul 1 11:14:06 PDT 2006 Kevin Karplus I haven't even looked at this, and it is due on Wed Jul 5. The HMMs like 2gvkA also (Evalue 3.5e-37) and nothing else. The alignment is essentially full-length, so the models probably won't need much fixing up. try1-opt2 looks ok, except for one strand. S178-H187 is predicted to be helix, not strand. Probably V164-G176 or thereabouts needs to be this strand. Sun Jul 2 11:58:03 PDT 2006 Kevin Karplus Now I'm real confused, as there doesn't seem to have ever *been* a try1-opt2. Perhaps I was looking at best-models.pdb.gz, which has only alignments in it. try1 crashed when it was first run. Let's do a real run of try1. Make started Sun Jul 2 12:01:21 PDT 2006 Running on lopez.cse.ucsc.edu Sun Jul 2 12:01:52 PDT 2006 Kevin Karplus Actually---I started a full make, so that the summary.html page would be updated also. Sun Jul 2 12:23:48 PDT 2006 Kevin Karplus We need to make a hexamer of T0328, but luckily the good hit 2gvkA is part of a hexamer, so we should be able to copy from that. Sun Jul 2 14:37:11 PDT 2006 Kevin Karplus I made 2gvkA/T0328-2gvkA.6mer-a2m in anticipation of needing to create the hexamer. Sun Jul 2 18:07:29 PDT 2006 Kevin Karplus try1-opt2 is very close to 2gvkA, but there are a few gaps that need closing. I'll do a polishing run before looking to see whether the hexamer is needed. Sun Jul 2 19:17:40 PDT 2006 Kevin Karplus Conformation[3] T0328.try1-opt2.pdb.gz has 35 breaks T0328.try1-opt2.pdb.gz breaks before (T0328)L146 with cost 4.12591 T0328.try1-opt2.pdb.gz breaks before (T0328)S237 with cost 3.53368 T0328.try1-opt2.pdb.gz breaks before (T0328)G279 with cost 3.36387 T0328.try1-opt2.pdb.gz breaks before (T0328)V32 with cost 3.09188 T0328.try1-opt2.pdb.gz breaks before (T0328)E135 with cost 2.91122 T0328.try1-opt2.pdb.gz breaks before (T0328)G254 with cost 1.37749 T0328.try1-opt2.pdb.gz breaks before (T0328)Q253 with cost 1.10189 T0328.try1-opt2.pdb.gz breaks before (T0328)E94 with cost 0.64002 Sun Jul 2 19:27:45 PDT 2006 Kevin Karplus try2 (a polishing run using the try1 models) started on shaw. Sun Jul 2 20:11:48 PDT 2006 Kevin Karplus try2 seems to be making very slow improvements. It might be a good idea to try starting from the gromacs models, which are close to the best scoring models in 3-space, but score poorly, so that undertaker can optimize more readily. Mon Jul 3 08:30:21 PDT 2006 Kevin Karplus Because PDB had only the monomer, and not the hexamer, I downloaded http://pqs.ebi.ac.uk/pqs-doc/macmol/2gvk.mmol and gzipped it as 2gvkA/2gvk.mmol.gz modifying make-hexamer1.under to use it. Mon Jul 3 08:43:05 PDT 2006 Kevin Karplus I started a short optimization run 6mer/try1 on cheep. Mon Jul 3 10:17:09 PDT 2006 Kevin Karplus Had to kill the run and start over---I hadn't finished editing 6mer/try1.under and had not removed the comment symbol in front of "include read-pdb.under", so we weren't starting from the 6mer! Mon Jul 3 12:52:50 PDT 2006 Kevin Karplus The 6mer run is going very slowly, and can't really handle the individual monomers anyway, since it is symmetry group S_{2,3}. I started a similar trimer run on shaw, just using 2gvkA ABC Mon Jul 3 19:52:31 PDT 2006 Kevin Karplus The 3mer run finished and seems to have improved the costfcn. The 6mer run is only at try1-opt1, and has actually made the scoring worse (perhaps DEF were not in the same order as ABC, so the "dimer" transform was not computed correctly?) Mon Jul 3 20:11:22 PDT 2006 Kevin Karplus The 6mer run is *definitely* wasting time. I'm going to kill it. Mon Jul 3 20:33:39 PDT 2006 Kevin Karplus Submitted with comment T0328 was modeled closely on 2gvkA. We attempted to optimize in a 6mer context, but undertaker does not understand the S_{3,2} symmetry group (only cyclic groups), so we had to try optimizing separately as a trimer (chains ABC) and a big dimer (ABC vs DEF). The "big dimer" run had the transformation between the two halves wrong (perhaps because of an incorrect naming choice for the monomers), and was terminated early. We got off to a late start on this target, so there was not time to rerun the "big dimer". Model 1 is try1-opt2 of the trimer run. Model 2 is try2-opt2, the monomer used as a basis for the trimer run. Model 3 is try2-opt2.repack-nonPC, the backbone of try2-opt2 repacked by rosetta. Model 4 is try1-opt2, the fully automatic model. Model 5 is sidechain replacement by SCWRL on an alignment to 2gvkA.