Fri May 9 10:10:02 PDT 2008 T0396 Make started Fri May 9 10:11:02 PDT 2008 Running on peep.cse.ucsc.edu Fri May 9 10:28:39 PDT 2008 Kevin Karplus dimer, pH 7.5, binds flavin adenine dinucleotide BLAST finds moderately close hit in 2hj3A Fri May 9 10:32:10 PDT 2008 Kevin Karplus Very clear conservation signal for some residues: W11 HF15 C46 C49 HYF78 NDS79 VLI81 NDST82 WYFLI99 Perhaps the C46,C49,H78 are a binding site? I wonder what the highly conserved W11 is for. HMM-based (AA-only) search gets a strong hit for the FAD-dependent thiol oxidase, which appeared in the iterated searches, slightly favoring 1oqcA over 2hj3A, so the BLAST hit seems to be the right one. It looks like there are 3 templates we can use: 1oqcA 2hj3A 1jr8A. (2hj3A does not appear in our template library). Fri May 9 14:39:07 PDT 2008 Kevin Karplus In try1-opt3, I don't like the way that W6 is oriented differently than in the templates. W11 has also moved a little, as has Y99. It might be worth adding a few constraints to pin them down a bit more. C46-C49 does form an SSbond, so that should be added as a known ssbond. Note: 2hj3A has 4 disulfides, but the target has only 2 cys. I should also include the 2hj3A template for try2. Fri May 9 17:47:47 PDT 2008 Kevin Karplus try2 looks pretty good, but I might want to redo as dimer. Mon May 19 22:36:07 PDT 2008 SAM-T08-MQAO hand QA T0396 Submitted Mon May 19 22:36:07 PDT 2008 SAM-T08-MQAU hand QA T0396 Submitted Mon May 19 22:36:07 PDT 2008 SAM-T08-MQAC hand QA T0396 Submitted Wed May 21 18:17:39 PDT 2008 Kevin Karplus MQAC likes the Zhang server, but MQAU prefers pipe_int and SAM-T08-server. Optimizing MQAU with try2.costfcn starts from SAM-T08-server_TS1, while optimizing MQAC starts from LEE-SERVER_TS5. The MQAU1-opt3 model scores best with try2.costfcn, primarily because it gets the align_constraints well and the ssbond, though all the models have decent ssbonds (except the gromacs or scwrl ones). MQAU1-opt3 unwinds the N-terminus a bit, which may match the alpha predictions a bit better. For try3, I'll polish from all existing models, with known_ssbond and align_constraints turned down a little, and soft_clashes and breaks turned up a little. Wed May 21 20:28:53 PDT 2008 Kevin Karplus The remaining differences between models is so small that I can't really see much point to further polishing. I doubt that I can predict the N- and C-termini any better than what I have currently in best-models.pdb.gz Question: should I try to optimize as a dimer?? Sun May 25 09:45:42 PDT 2008 Kevin Karplus Rescoring models with try3.costfcn after changing the clash definition. try3-opt2 has a few more clashes that I would like, so I'll do one more polishing run with clashes turned up higher, before trying to make a dimer based on 1oqcA and 1oqcC. Sun May 25 09:57:32 PDT 2008 Kevin Karplus try4 started on shaw. Sun May 25 16:44:56 PDT 2008 Kevin Karplus try4-opt3 scores best with the try4 costcn, but rosetta prefers try3-opt3.gromacs0.repack-nonPC slighltly. Sun May 25 17:06:28 PDT 2008 Kevin Karplus I made a dimer out of try4-opt3, using 1oqcA and 1oqcC as a model (following the instructions in casp8/README). Using the same try4.costfcn as before, I'll optimize the dimer model. I expect that this will tweak the multimer a bit and pack things a little tighter at the interface. Mon May 26 04:06:41 PDT 2008 Kevin Karplus Rats! undertaker failed on dimer/try4 with a rather uninformative assertion failure. It got as far as try4-opt2 before crashing, and I can see one problem: it is moving the two monomers apart to relieve clashes, rather than repacking sidechains. I should try another run without TweakMultimer. I also see that I omitted the KnownBreak command from dimer/costfcn-init.under I remade dimer/costfcn-init.under and it seems to be ok now. I'll modify dimer/try5 to be like dimer/try4, but with no TweakMultimer and with the Rotamer operations turned up high. Mon May 26 05:58:52 PDT 2008 Kevin Karplus dimer/try5-opt3 looks better, though it could still be better packed in the interface. I'll turn up the dry terms, phobic_fit, and pred_nb11_back even higher, and turn on TweakMultimer to see if we can pack things any tighter---probably won't be successful. Mon May 26 08:15:37 PDT 2008 Kevin Karplus dimer/try6-opt3 scores better and looks pretty uniformly dense, The differences between the models from try1 through try6 are all pretty small though, so I'll declare this model done: ReadConformPDB from-dimer-try6-opt3.pdb < try3-opt3 < MQAU1