Tue Jul 1 09:25:17 PDT 2008 T0480 Make started Tue Jul 1 09:25:42 PDT 2008 Running on cheep.cse.ucsc.edu Tue Jul 1 09:27:37 PDT 2008 Kevin Karplus very short (55 residues) NMR model. PDB blast finds nothing (best is 2zkr2 at E-value 7.7) Tue Jul 1 09:59:38 PDT 2008 Kevin Karplus I seem to be getting lots of hits to ubiquitin/ribosomal protein S27a in the t06 multiple alignment. The single-track HMMS seem to like the 2zkr2 template best also, though with E-value 2. It isn't in the template library, so I may need to add it as an extra hit. 2ausB is coming up as the top hit in the t06 template library. There seem to be some conserved CYS residues: C21, C24, C39, C42 This makes me suspect a zinc binding site. Tue Jul 1 10:14:54 PDT 2008 Kevin Karplus The reason we don't have 2zkr2 in the template library is that it is a complex obtained by docking homology models into an 8.7 Angstrom cryo-EM map, so it is only homology modeling at the atomic level anyway. This is likely to be a zinc-finger, which shouldn't be too hard to model. Here is whet NCBI has to say about the conserved domain: cl00876: Ribosomal_S27 Super-family Ribosomal protein S27a. This family of ribosomal proteins consists mainly of the 40S ribosomal protein S27a which is synthesised as a C-terminal extension of ubiquitin (CEP). The S27a domain compromises the C-terminal half of the protein. The synthesis of ribosomal proteins as extensions of ubiquitin promotes their incorporation into nascent ribosomes by a transient metabolic stabilisation and is required for efficient ribosome biogenesis. The ribosomal extension protein S27a contains a basic region that is proposed to form a zinc finger; its fusion gene is proposed as a mechanism to maintain a fixed ratio between ubiquitin necessary for degrading proteins and ribosomes a source of proteins. I believe that the text is originally from PFAM01599. Tue Jul 1 10:46:34 PDT 2008 Kevin Karplus I think that the top 3 hits (2ausB, 2apoB, 1rlyA) are all relevant ribosome biogenesis proteins. In 2ausB, the SG atoms are within 3.4-4.5 Angstroms of each other. The maybe_metal goal of 3.6 Angstroms seems pretty good. The CB atoms of the CYS residues vary more, as the symmetry breaks down further from the zinc. The 2apoB model has two different conformations for the CYS residues around the zinc, and neither looks great. The 1rlyA model has only 3 CYS, not 4, and is a NMR model with very floppy tails. I think we want to concentrate on 2ausB and 2apoB, though other g.41.* templates may also be relevant. The protein is not compact in the templates, so there isn't much reason to expect it to be in the NMR structure for T0480, except that they seem to have trimmed the protein down to just the compact core. For try2, I'll probably want to turn maybe_metal way up and eliminate maybe_ssbond. Tue Jul 1 11:35:06 PDT 2008 Kevin Karplus try2 started to polish try1 models with constraints on the SG atoms *and* maybe_metal turned up. If this doesn't succeed in undoing the erroneous disulfides introduced in try1, then I'll run again from alignments. Maybe I should do that anyway, to see what I get. Tue Jul 1 13:33:39 PDT 2008 Kevin Karplus I think I like the N-terminus of try3 better than try2, but the C-terminus of try2 better. I'm tempted to try to create another sheet constraint, with K13-R19 antiparallel to E32-G27 (note: I still need to figure out the phase). Tue Jul 1 13:55:53 PDT 2008 Kevin Karplus I'll start another run (try4) from alignments, encouraging try3 sheets a bit. I probably want to do some more runs from alignments different sheet constraints, to see what I can come up with. Tue Jul 1 15:23:10 PDT 2008 Kevin Karplus try4 winds some of the predicted strands into helices. I think I'll try adding some sheet constraints of my own. try5 will have SheetConstraint R36 C39 F47 Y44 hbond Y37 10 SheetConstraint G27 A31 G40 R36 hbond F29 10 SheetConstraint R16 P22 A31 G25 Hbond G25 2 SheetConstraint Y7 A10 R16 K13 Hbond K13 2 and will look for these sheets from the alignments. Sun Jul 6 08:57:59 PDT 2008 Kevin Karplus The MQAC quality assessment favors pro-sp3-TASSER_TS2, MULTICOM-RANK_TS5, SAM-T08-server_TS1, pro-sp3-TASSER_TS1, MULTICOM-REFINE_TS5, ... but the top predicted GDT is only 49%. The MQAU quality assessment like the SAM-T08-server and Zhang-Server models, but with top GDT predicted to be only 35%. I'll do MQAU1, MQAC1, MQAU5, and MQAC5 metaserver runs. Sun Jul 6 09:12:39 PDT 2008 Kevin Karplus The 2ausB alignment is far from complete, and does not include C39 or C42. The same is true for the 2apoB alignment. The 1rlyA alignment suggests a zinc-binding site, though C21 and C39 are clashing The 2ct7A alignment also suggests a zinc-binding site. The 2e72A alignment, like 2ausB and 2apoB, does not have C39 or C41. The 2bmoA alignment suggests a C21-C39 C24-C42 pair of disulfides, though it could be compressed down to a metal binding site. I like the sheets in 2bmoA, so maybe I should try making a costfcn with those sheets and a zinc-binding site. Let's see what the meta-servers come up with first. Mon Jul 7 13:56:47 PDT 2008 Kevin Karplus Top models for different costfcns try1 MQAU1-opt3 MQAC1-opt3 MQAC5-opt3 try5-opt3 try1-opt3 try2-opt3 try2 MQAU1-opt3 MQAC1-opt3 try2-opt3 try3-opt3 try1-opt3 MQAU5-opt3 MQAC5-opt3.gromacs0 try3 MQAU1-opt3 MQAC1-opt3 try2-opt3 try3-opt3 try1-opt3 MQAU5-opt3 MQAC5-opt3.gromacs0 try4 MQAU5-opt3 try5-opt3 MQAC5-opt3 try4-opt3 MQAU1-opt3.gromacs0 MQAC1-opt3.gromacs0 try2-opt3.gromacs0 try5 try5-opt3 MQAU5-opt3 MQAC5-opt3 try4-opt3 MQAU1-opt3.gromacs0 MQAC1-opt3.gromacs0 try2-opt3.gromacs0 Mon Jul 7 14:10:33 PDT 2008 Kevin Karplus MQAU5-opt3 makes a flat sheet but has a decent zinc site. try5-opt3 has a 3-strand sheet and a 2-strand sheet and a decent zinc site. MQAC5-opt3 has a 3-strand sheet and a 2-strand sheet and a decent zinc site, but the hairpin exposes L14 and Y7 which are predicted to be buried. try4-opt3 has a 3-strand sheet and a decent zinc site. MQAU1-opt3 is similar to MQAC5, but with a collapsed zinc site MQAC1-opt3 is similar to MQAC5, but with an incompletely formed zinc site try2-opt3 forms a 3-strand sheet and the C21-C39 disulfide. try3-opt3 is similar to try2-opt3, but has another hairpin Thu Jul 10 10:57:49 PDT 2008 SAM-T08-MQAO hand QA T0480 Submitted Thu Jul 10 10:57:49 PDT 2008 SAM-T08-MQAU hand QA T0480 Submitted Thu Jul 10 10:57:49 PDT 2008 SAM-T08-MQAC hand QA T0480 Submitted Thu Jul 24 15:47:57 PDT 2008 Kevin Karplus Looking at this for the first time in over two weeks, I noticed that the zinc site in 2ct7A is quite similar (in terms of 2ry structure) to what I expect to see here, but that the hairpin around the first CYS pair and the the hairpin around the second CYS pair are not part of the same sheet. Maybe I should try making a model with just 2ct7A as alignments. Thu Jul 24 15:58:34 PDT 2008 Kevin Karplus I started a try6 run with just 2ct7A as an alignment. The rest will have to be built from short fragments, which may not work. Initially the try6 costfcn likes MQAU5-opt3, which has the best zinc-binding site so far. Thu Jul 24 19:36:46 PDT 2008 Kevin Karplus The try6 model does get the 2ct7A version of the zinc site pretty well. None of the other models I've looked at have quite the same relationship between the hairpins, generally going for the extra Hbonds of a more complete sheet. I'm interested in trying a chimera: try6 for the middle and MQAC5 for the ends. Say M1-K17,T45-S55 from MQAC5-opt3, H18-Y44 from try6-opt3 Thu Jul 24 21:08:14 PDT 2008 Kevin Karplus The try7 model does not seem to be much of an improvement over the try6 model---it may even be slightly worse. For try8, I'm repeating try6, but with 2bmoA as the sole template, rather than 2ct7A. Thu Jul 24 21:50:01 PDT 2008 Kevin Karplus I rather like try8, and will try polishing it in try9 to close breaks and remove clashes. Fri Jul 25 04:20:05 PDT 2008 Kevin Karplus try9 does quite well---I'll probably include it as one of my final models. I'll do a polishing run as try10 which turns up the burial and predicted local-structure terms, and which has no sheet constraints. I expect that this will mainly try to polish MQAU5, so for try11, I'll use the same costfcn, but exclude MQAU5 models from the optimization. Fri Jul 25 04:45:30 PDT 2008 Kevin Karplus try10 does indeed polish MQAU5-opt3. Rosetta still prefers MQAU5-opt3.gromacs0.repack-nonPC try11 polishes try9-opt3 I have got at least 5 models that agree fairly well on the zinc site, but get more variable further away. I think I've reached the point of diminishing returns on this target and will submit. I've decided to omit the try6 and try7 models (from 2ct7A) although they have good zinc sites, because the rest of the model is rather sloppy looking. Fri Jul 25 05:18:08 PDT 2008 Kevin Karplus Submitted with comment: The most distinctive feature of T0480 were the conserved CYS residues: C21, C24, C39, and C42. I decided to make these into a tetrahedral zinc-binding site. Most of my models came from SAM+undertaker alone, but the first model was optimized from a meta-server model. I'm a bit dubious about the flat sheet in that model, but it consistently scored best in several variants of the cost function. Model 1 T0480.try10-opt3.pdb # < MQAU5-opt3 < RBO-Proteus_TS4 2 T0480.try11-opt3.pdb # < try9-opt3 < try8-opt3.gromacs0 < align(2bmoA) 3 T0480.try5-opt3.pdb # < align(2hf1A) 4 T0480.try4-opt3.pdb # < align(1v6kA) 5 T0480.MQAC5-opt3.gromacs0.pdb # < SAM-T08-server_TS1