Tue Jun 17 09:34:17 PDT 2008 T0457 Make started Tue Jun 17 09:35:02 PDT 2008 Running on peep.cse.ucsc.edu Tue Jun 17 10:43:45 PDT 2008 Kevin Karplus The t06 multiple alignment contains 11 sequences of PDB files: 1ir6A, 1wpmA, 1wppA, 1i74A, 1k20A, 1k23A, 2enxA, 2iw4A, 2eb0A, 1wpnA, 2qb6A all phosphatases. There is most likely a manganese binding site in the target. I may want to find it and use some scaffolding to preserve it. The t04 multiple alignment also has the PDB files, but the less sensitive t2k search does not include them. The sequences in t2k seem to be mainly phosphoesterases. Tue Jun 17 12:50:42 PDT 2008 Kevin Karplus The t06 HMMs find 5 templates (c.107.1.*): 1ir6A 1k20A 2eb0A 2qb7A 1wpnA When searching all (nonredundant) sequencecs of pdb, then 10 templates come up: 1ir6A 2hawA 2iw4A 1i74A 1wppA 1k20A 2enxA 1wpnA 2eb0A 2qb7A The t04 HMMs seemt od about the same as the t06 ones, and the t2k w0.5 search gets the same top 10 hits, though 2qb7A's E-value is not strong good (about 0.1). The t2k 2-track HMM with str2 gets 6 hits: 1k20A 1i74A 2eb0A 1wpnA 1ir6A 2qb7A Note: none of these hits came up in the BLAST search, so this is a fold-recognition target, not a trivial homology target. Tue Jun 17 15:14:04 PDT 2008 Kevin Karplus The try1 run seems to have started with 1i74A, but then added 2z98A. The optimization ignores the later addition though, and optimizes the the first alignment to 1i74A. Tue Jun 17 23:54:15 PDT 2008 Kevin Karplus clearly more break and clash removal is needed, as the best-scoring model with the try1 costfcn is try1-opt3.gromacs0.repack-nonPC, even though that really messes up the N-CA-C bond angles. There is a cluster of conserved HIS and ASP residues: H23,H106,H107 D27,D29,D83,D105,D158 I should look how they are arranged in the alignments and add appropriate constraints to hold them in place in the absence of the metal ion. There is a big hydrophobic spot on the monomer--do I have to dimerize this protein? Sat Jun 28 14:40:51 PDT 2008 SAM-T08-MQAO hand QA T0457 Submitted Sat Jun 28 14:40:51 PDT 2008 SAM-T08-MQAU hand QA T0457 Submitted Sat Jun 28 14:40:51 PDT 2008 SAM-T08-MQAC hand QA T0457 Submitted Wed Jul 2 13:35:23 PDT 2008 Kevin Karplus I started the metaserver runs today---I've not worked on this protein since first day, which is sort of embarrasing. Oh well, at least I have a few days left to look at it. The top hit (1k20A) is a dimer, and the sheet continues across the dimerization interface, so optimizing as a dimer is probably worthwhile. Wed Jul 2 14:46:48 PDT 2008 Kevin Karplus MQAU1 is based on SAM-T08-server as expected, but MQAC1 is based on pro-sp3-TASSER_TS4 Wed Jul 2 19:12:26 PDT 2008 Kevin Karplus top scoring with try1 is MQAU1-opt3.gromacs0, then MQAC1-opt3, then try1-opt3.gromacs0.repack-nonPC I'll make a dimer of MQAU1-opt3.gromacs0 with the 1k20 template, and try optimizing it. I'll probably want to add constraints for the cluster of conserved residues H23, D27, D29, S33, D83, H106, H107, D158 but I won't do it for try2. Wed Jul 2 19:44:06 PDT 2008 Kevin Karplus dimer/try2 started on peep. Thu Jul 3 04:16:47 PDT 2008 Kevin Karplus The dimer/try2 model is messed up at the dimer interface, with the oxygen atoms touching. Perhaps I have the constraints wrong? No---the SheetConstraint in try2 seems correct. Perhaps I have to make the dimer constraint stronger. The N-terminal helix seems to be bumping a bit, so perhaps I need to build a dimer with a different monomer also: I'll try both MQAC1-opt3 and try1-opt3.gromacs0.repack-nonPC Thu Jul 3 04:44:21 PDT 2008 Kevin Karplus I started dimer/try3 on peep, optiizing from three models built on 1k20 (not using the dimer/try2 models, so as not to get trapped in the same local optimum). If this doesn't work, I can try building dimers from 2eb0A or 1ir6A. Thu Jul 3 09:43:29 PDT 2008 Kevin Karplus The dimer/try3 model looks fine, so I don't need to build dimers based on other templates. The biggest remaining question is the phase of the strand around Q220-D226, as that is different in the models of different lineage. I'm not seeing much in the way of conservation or residue-residue contact prediction to guide the choice of phase. I wrote earlier I'll probably want to add constraints for the cluster of conserved residues H23, D27, D29, S33, D83, H106, H107, D158 Let me try to figure out the correspondence in 1k20A for the residues coordinating the Mn ions or SO4 group. restrict within(4.1, atomno=4710 or atomno=4709 or atomno=4713) and not HOH and not carbon T0457 1k20A H23 H9 D27 D13 D29 D15 A30 A16 D83 D77 H106 H99 H107 H100 T126 A121 D158 D151 ? K207 alignment uncertain ? R297 alignment uncertain ? K298 alignment uncertain Distances H23.NE2 D27.OD1 4.06 H23.NE2 D27.OD2 3.69 H23.NE2 D83.OD1 3.43 H23.NE2 A20.N 5.19 D27.OD1 D29.OD2 5.96 D27.OD1 H23.NE2 4.06 D27.OD1 A30.N 4.90 D27.OD2 D29.OD2 5.44 D27.OD2 A30.N 2.90 D27.OD2 H23.NE2 3.69 D83.OD1 D29.OD2 3.57 D83.OD1 A30.N 3.19 D83.OD1 D27.OD2 3.28 D83.OD1 D27.OD1 4.20 D83.OD2 T126.N 4.45 D83.OD2 D29.OD2 3.05 D83.OD2 H106.NE2 2.18 D83.OD2 D158.OD2 4.32 H106.NE2 T126.N 3.57 H106.NE2 D83.OD2 2.82 H106.NE2 D158.OD2 3.40 H107.NE2 D83.OD2 4.18 H107.NE2 H106.ND1 4.62 H107.NE2 H106.NE2 3.80 D158.OD1 D27.OD1 3.65 D158.OD2 D29.OD2 2.95 D158.OD2 D83.OD2 4.32 Thu Jul 3 10:53:38 PDT 2008 Kevin Karplus I added the "Mn" constraints to try4 and will try optimizing the monomer (including monomers from the try2 and try3 dimers). After that, I'll do the dimer optimization again. Thu Jul 3 13:22:24 PDT 2008 Kevin Karplus The try4 run did not start from the dimer-optimized monomers, but from MQAU1-opt3.repack-nonPC. Before I dimerize and optimize the dimer, let me do an "MQAX4" run that uses the try4 costfcn starting from all models EXCEPT SAM models. Thu Jul 3 15:10:41 PDT 2008 Kevin Karplus The MQAX4 model is being built mainly from BAKER-ROBETTA_TS3 Fri Jul 4 09:06:33 PDT 2008 Kevin Karplus MQAX4-opt3.gromacs0 scores a little worse than try4, but better than the earlier tries with the try4 costfcn. The try1 costfcn likes try4 best, then MQAU1-opt3.gromacs0, then MQAX4-opt3.gromacs0. Rosetta likes best MQAX4-opt3.gromacs0.repack-nonPC The try4 and MQAX4 models are very similar, but try4 packs the two domains together a little tighter (D214-P88 is 6.24Ang instead of 8.33Ang). I think this is almost done---I'll make dimers from the try4 and MQAX4 models, based on 1k20 and 2eb0. Fri Jul 4 09:41:29 PDT 2008 Kevin Karplus dimer/try5 started on cheep, to optimize dimers made from try4 and MQAX4 based on 1k20 and 2eb0. I don't expect a lot of changes (other than fixing the dimer transform), so the try5 chain A will probably be my first choice, followed by as different chains as I can find in my set of models. Fri Jul 4 13:38:35 PDT 2008 Kevin Karplus dimer/try5-opt3 scores well, and rosetta likes T0457.try5-opt3.unpack.gromacs0.repack-nonPC.pdb.gz :totals -2164.3 785.9 1169.5 0.0 -35.8 649.1 1.1 -44.6 -57.1 119.1 0.0 0.0 0.0 0.0 -264.3 Fri Jul 4 13:53:34 PDT 2008 Kevin Karplus I think I'm done with this target. I'll submit the dimer-optimized models first: InFilePrefix dimer/decoys/ ReadConformPDB T0457.try5-opt3.unpack.gromacs0.repack-nonPC.pdb chain A # < dimer-try4-2eb0 InFilePrefix dimer/decoys/ ReadConformPDB T0457.try3-opt3.unpack.gromacs0.pdb chain A # < dimer-MQAU1-1k20 InFilePrefix decoys/ ReadConformPDB T0457.MQAX4-opt3.gromacs0.pdb # < BAKER-ROBETTA_TS3 InFilePrefix decoys/ ReadConformPDB T0457.try1-opt3.gromacs0.repack-nonPC.pdb # < align(1i74A) InFilePrefix decoys/ ReadConformPDB T0457.MQAC1-opt3.pdb # < pro-sp3-TASSER_TS4