Thu Jun 15 09:10:51 PDT 2006 T0332 Make started Thu Jun 15 09:11:41 PDT 2006 Running on lopez.cse.ucsc.edu Thu Jun 15 09:16:49 PDT 2006 Kevin Karplus Strong blast hit to 1zjrA (24%id over 156 residues, 6.6e-12), 1ipaA (32%id over 152 residues, 5.6e-11) 1v2xA, 1x7oA Moderate blast hits to 1gz0A and 1mxiA. Thu Jun 15 18:11:19 PDT 2006 Kevin Karplus Strong HMM hits to 1v2xA, 1x7oA, ... (c.116.1.*) All the top alignments are in agreement with the try1-opt2 model. There is a large hydrophobic patch which probably indicates a dimer. We should probably optimize this as a dimer. Mon Jul 3 10:23:28 PDT 2006 Kevin Karplus I'll use 1x7o[AB] as the model for the dimer. No, that has a large other domain that may cause problems. I'll pick up http://pqs.ebi.ac.uk/pqs-doc/macmol/1v2x.mmol instead and use it (1v2x relies on symmetry for the other monomer). Mon Jul 3 10:38:01 PDT 2006 Kevin Karplus Made dimer/decoys/dimer-1v2xA-from-try1.pdb using undertaker < make-dimer1.under Mon Jul 3 10:47:16 PDT 2006 Kevin Karplus Trying a simple optimization in dimer/try1 (on lopez) If it works ok, I'll do a polishing run with breaks and clashes up and constraints off. Mon Jul 3 17:06:30 PDT 2006 Kevin Karplus The dimer/try1 run looks fairly good, though a little bit foamy still. I'll turn up the packing terms for try2. Mon Jul 3 17:15:37 PDT 2006 Kevin Karplus dimer/try2 with stronger packing terms and no constraints started on farm cluster Tue Jul 4 11:13:51 PDT 2006 Kevin Karplus For some reason gromacs runs seem to fail on the farm cluster. Perhaps there is a missing library? I'm remaking dimer/ T0332.do2, so that the gromacs file for try2-opt2 will be created. Tue Jul 4 11:16:35 PDT 2006 Kevin Karplus oops---no real point to that, since gromacs doesn't understand big chain breaks in a multimer. Can it handle unpacked multimers, with properly separate chains? Tue Jul 4 11:24:16 PDT 2006 Kevin Karplus Yes, that is the way to run gromacs on multimers: make decoys/T0332.try2-opt2.unpack.pdb.gz decoys/T0332.try2-opt2.unpack.gromacs0.pdb.gz The differences between dimer/ try1 and dimer/try2 are tiny and try2-opt2 still has a couple of bad breaks: T0332.try2-opt2.pdb.gz breaks before (T0332)S45 with cost 2.42046 T0332.try2-opt2.pdb.gz breaks before (T0332)I44 with cost 1.72478 I think that I44.N-Q42.O is trying to form an inappropriate Hbond. (No, undertaker properly sees that as a clash, not an Hbond.) Perhaps I should optimize the monomer a bit more before dimerizing. Tue Jul 4 11:46:22 PDT 2006 Kevin Karplus try2 started on cheep. I'm optimizing just from the gromacs models, since I wanted to fix a bad clash. I am also giving the run sidechain repacking by rosetta and scwrl on the gromacs backbone, since gromacs makes such a hash of sidechains. Tue Jul 4 12:42:05 PDT 2006 Kevin Karplus try2-opt2 certainly reduced breaks and clashes. It scores best with unconstrained, try2, and try1 costfcns. Rosetta likes best repacking try2-opt2.gromacs0 I'll do another polishing run with breaks and clashes turned up higher and contraints off. I'll also start from just the gromacs0.repack-nonPC models. Tue Jul 4 12:49:59 PDT 2006 Kevin Karplus try3 started on cheep. Tue Jul 4 13:38:46 PDT 2006 Kevin Karplus try3-opt2 did not beat try2-opt2 on the try3 costfcn, but rosetta likes best decoys/T0332.try3-opt2.gromacs0.repack-nonPC.pdb.gz Tue Jul 4 13:40:58 PDT 2006 Kevin Karplus I used make-dimer2.under and make-dimer3.under to make dimers from try2-opt2 and try3-opt2.gromacs0.repack-nonPC Tue Jul 4 14:06:57 PDT 2006 Kevin Karplus Starting dimer/try3 on cheep, which will polish the dimer models that were built directly from monomers (I expect it will work mainly with dimer-1v2xA-from-try2.pdb). Tue Jul 4 16:22:02 PDT 2006 Kevin Karplus dimer/try3-opt2 scores best with the dimer/try3.costfcn, dimer/try2-opt2 scores best with the dimer/try2.costfcn and dimer/try1.costfcn rosetta likes best decoys/T0332.try3-opt2.unpack.gromacs0.repack-nonPC.pdb.gz I actually like the way that rosetta folds Y232 in, so I'll make decoys/T0332.try3-opt2.unpack.gromacs0.repack-nonPC.pdb.gz be my first choice. Tue Jul 4 16:47:27 PDT 2006 Kevin Karplus Monomers submitted with comment We did some optimization in a dimer context (based on 1v2x[AB]), though very little change was made---mainly tweaking the dimer transform, not changing the monomers much. Model 1 is chain A from dimer/ try3-opt2.unpack.gromacs0.repack-nonPC the dimer that rosetta liked best after repacking. Model 2 is chain A from dimer/ try3-opt2 the dimer that undertaker liked best. Model 3 is try2-opt2, the optimization in monomeric context that scored best with undertaker. Model 4 is try3-opt2, another optimization in monomeric context that scored almost the same as try2-opt2. Model 5 is sidechain replacement by SCWRL on an alignment to 1v2xA. Dimers submitted with comment We did some optimization in a dimer context (based on 1v2x[AB]), though very little change was made---mainly tweaking the dimer transform, not changing the monomers much. Model 1 is try3-opt2.unpack.gromacs0.repack-nonPC, the dimer that rosetta liked best after repacking the sidechains. Model 2 is try3-opt2, the best-scoring dimer with undertaker. Model 3 is try2-opt2, polished from try1-opt2. Model 4 is try1-opt2, based on a dimer using the automatic model from the monomer optimization. Wed Jul 12 17:33:53 PDT 2006 Sylvia Do There were some parts for try3-op2 that I wanted to see packed closer. try4 will be try3-op2 with these changes to the costfcn: From SetCost wet6.5 15 near_backbone 5 way_back 5 dry5 15 dry6.5 20 dry8 15 dry12 5 to SetCost wet6.5 15 near_backbone 5 way_back 5 dry5 15 dry6.5 40 dry8 20 dry12 5 I increased the weights for dry6.5 and dry 8 for try4. try4 is running on Heehaw Thu Jul 13 14:34:30 PDT 2006 Sylvia Do It looks like try2 and try 3 both beat try4 with try4's costfcn.