Thu Jul 10 09:22:57 PDT 2008 T0496 Make started Thu Jul 10 09:23:26 PDT 2008 Running on cheep.cse.ucsc.edu Thu Jul 10 12:43:37 PDT 2008 Kevin Karplus This looks like a template-free prediction---the best E-value is 41.6 for 2f9iB. Thu Jul 10 13:56:55 PDT 2008 Kevin Karplus try1 gets a good 3-strand mixed sheet, which none of the top models in undetaker-align have. I should find out from VAST where it comes from. There are currently 3 separate chunks: a helical region M1-T172 the mixed sheet T172-A159 a helical region A159-L178 Your VAST Search job was submitted at 07/10/2008 17:01:44(EDT). Request ID: 1078132771598667957 I think the C-terminal helices for a 3-helix bundle, and the N-terminal ones may also, though Q40-Y53 probably packs against the sheet. The 3-strand sheet almost certainly comes from the top hit (2f9iB), but not from the str2+near-backbone-11 alignment. VAST finds a few possibilities for the first domain (1womA, 1va4A, 1absA, 1a88A, 1zq1C, 2vqzD), but none of these seem to be in the top hits, though some similar structue may be. 1vl9A and 1n28A may span the boundary between the first two domains, though neither was used as a template. VAST finds no similarties for the 3rd domain, so it is quite likely still badly misfolded. Except for K10 and D157, the conserved residues look like hydrophobic core residues. For try2, I'll put in strand and sheet constraints from try1, and try to close gaps a bit. For try3, I'll use the same costfcn, but run from alignments. Thu Jul 10 21:08:00 PDT 2008 Kevin Karplus For T0496, it looks like starting from try1 and reoptimizing (try2-opt3) worked better than starting over from alignments (try3-opt3.gromasc0). The two models agree on Q25-N119, but try2 has better packing of both the N-terminal and C-terminal helices. try3 may have been trying a little too hard to close gaps, so I'll do another run from alignments with a smaller break cost. Fri Jul 11 00:17:29 PDT 2008 Kevin Karplus try4-opt3 beats try3-opt3.repack-nonPC, but not try2-opt3 Fri Jul 11 00:23:12 PDT 2008 Kevin Karplus Some relationships seem preserved across all the models so far, but the helices at the beginning and end of the chain are causing me some difficulty. Mon Jul 14 09:17:03 PDT 2008 Kevin Karplus Both the MQAU and MQAC quality assessments like SAM-T08-server best, but the predicted GDTs are quite low (around 30%). I've started MQAC1 and MQAU1 runs, but I probably should start runs excluding the SAM servers also. Wed Jul 16 19:34:24 PDT 2008 Kevin Karplus Indeed, the MQAU1 and MQAC1 runs both polish SAM-T08-server_TS1, but MQAY1 polishes Pcons_dot_net_TS5. costfcn favorite models try1 try2-opt3 try1.opt3 MQAC1-opt3 MQAU1-opt3 try3-opt3.repack-nonPC try4-opt3 MQAY1-opt3 try2 try2-opt3 try3-opt3.gromacs0 try1-opt3.gromacs0 try4-opt3 MQAY1-opt3.gromacs0 MQAC1-opt3.gromacs0 MQAU1-opt3 try3 try2-opt3 try3-opt3.gromacs0 try1-opt3.gromacs0 try4-opt3 MQAY1-opt3.gromacs0 MQAC1-opt3.gromacs0 MQAU1-opt3 try4 try2-opt3 try4-opt3 try3-opt3.repack-nonPC try1-opt3.gromacs0.repack-nonPC MQAY1-opt3 MQAC1-opt3 MQAU1-opt3 rosetta MQAY1 try4 try2 try3 try1 MQAC1 MQAU1 (all -opt3.gromacs0.repack-nonPC) Wed Jul 16 19:46:57 PDT 2008 Kevin Karplus The MQAY1 model (from Pcons_dot_net, and probably before that from BAKER-ROBETTA) has a 4-strand sheet, rather than the 3-strand sheet that the rest have. The C-terminal helices have been folded more than in the other models, and everything after S106 seems to be treated independently. How consistent are the sheets? MQAC1-opt3:SheetConstraint (T0496)D96 (T0496)Q100 (T0496)F70 (T0496)K74 hbond (T0496)Q100 1 MQAC1-opt3:SheetConstraint (T0496)G57 (T0496)V65 (T0496)N76 (T0496)T68 hbond (T0496)V59 1 MQAU1-opt3:SheetConstraint (T0496)D96 (T0496)Q100 (T0496)F70 (T0496)K74 hbond (T0496)Q100 1 MQAU1-opt3:SheetConstraint (T0496)G57 (T0496)V65 (T0496)N76 (T0496)T68 hbond (T0496)V59 1 MQAY1-opt3:SheetConstraint (T0496)P55 (T0496)A62 (T0496)E95 (T0496)N102 hbond (T0496)S61 1 MQAY1-opt3:SheetConstraint (T0496)V59 (T0496)C64 (T0496)F73 (T0496)T68 hbond (T0496)A62 1 try1-opt3: SheetConstraint (T0496)D96 (T0496)Q100 (T0496)F70 (T0496)K74 hbond (T0496)Q100 1 try1-opt3: SheetConstraint (T0496)G57 (T0496)V65 (T0496)N76 (T0496)T68 hbond (T0496)V59 1 try2-opt3: SheetConstraint (T0496)I97 (T0496)Q100 (T0496)H71 (T0496)K74 hbond (T0496)Q100 1 try2-opt3: SheetConstraint (T0496)G57 (T0496)V65 (T0496)N76 (T0496)T68 hbond (T0496)V59 1 try3-opt3: SheetConstraint (T0496)I97 (T0496)L101 (T0496)H71 (T0496)Q75 hbond (T0496)Q100 1 try3-opt3: SheetConstraint (T0496)G57 (T0496)V65 (T0496)N76 (T0496)T68 hbond (T0496)V59 1 try4-opt3: SheetConstraint (T0496)D96 (T0496)L101 (T0496)F70 (T0496)Q75 hbond (T0496)Q100 1 try4-opt3: SheetConstraint (T0496)G57 (T0496)V65 (T0496)N76 (T0496)T68 hbond (T0496)V59 1 The MQAY1 model has a very unusual spearation 4 hairpin, and puts the E95-N102 strand on the other side of the hairpin. The packing of the strands looks interesting, but I don't believe the sep=4 hairpin. Thu Jul 17 16:58:00 PDT 2008 SAM-T08-MQAO hand QA T0496 Submitted Thu Jul 17 16:58:00 PDT 2008 SAM-T08-MQAU hand QA T0496 Submitted Thu Jul 17 16:58:00 PDT 2008 SAM-T08-MQAC hand QA T0496 Submitted Fri Jul 25 13:03:14 PDT 2008 Kevin Karplus I should probably redo the extraction of sheets from MQAY1, as only 3 of the 4 strands are currently extracted, and undertaker was changed fairly recently to be more loose in deciding when something was a strand. Fri Jul 25 13:08:39 PDT 2008 Kevin Karplus OK, extracting from MQAY1-opt3.gromacs0.repack-nonPC gives me SheetConstraint (T0496)E95 (T0496)N102 (T0496)P55 (T0496)A62 hbond (T0496)N102 1 SheetConstraint (T0496)F70 (T0496)F73 (T0496)T80 (T0496)V77 hbond (T0496)H71 1 SheetConstraint (T0496)V59 (T0496)C64 (T0496)F73 (T0496)T68 hbond (T0496)M60 1 which picks up the not-quite-formed strand for V77-T80 Mon Jul 28 14:06:21 PDT 2008 Kevin Karplus I'll do a polishing run for MQAY1-opt3 and one for try2/3/4 try5 started to polish MQAY1, try6 to polish try2/3/4 The sheets in these two lineages are quite different: try2/3/4: 1 ^v 2 || 4 MQAY1: 4 || 1 ^v 2 ^v 3 though I might be able to do cut-and-paste for the C-terminal helices, after the last strand of the sheet to mix-and-match domains. Mon Jul 28 16:35:03 PDT 2008 Kevin Karplus try6 and try5 improve on try4 and MQAY1 respectively, and rosetta's new favorite it try5-opt3.gromacs0.repack-nonPC I think I'd like to make a try6-try3 chimera, crossing over around N119-P120, with try6 for the N-terminal end, and try3 for the C-terminal end. Mon Jul 28 16:47:37 PDT 2008 Kevin Karplus try7 started to optimize chimera-try6-try3. Mon Jul 28 20:43:50 PDT 2008 Kevin Karplus try7 has new top score in the try1,2,3,4,6 line. Rosetta likes it, but not as much as try5. Looking at the models in spacefill mode, I rather like the try5 model. The try7 one is too loose. try3 and try4 might be ok, if the sheet edge is part of a dimer interface. I should pass both try5 and try7 to VAST, to see what they are related to, and whether those folds dimerize. try5 search: Your VAST Search job was submitted at 07/28/2008 23:54:22(EDT). Request ID: 388126691340854357 For the whole chain, the segment that gets aligned is about 31-120. VAST sees a domain break around 120 (not too surprising, since I did some mix-and-match with a break right there). Looking at some of the templates that match before P120 might help me find something to model the whole protein on. PDB C D Ali. Len. SCORE P-VAL RMSD %Id Description 1FBV A 3 40 7.8 10e-4.8 2.3 7.5 Structure Of A Cbl-Ubch7 Complex: Ring Domain Function In Ubiquitin-Protein Ligases 1JWO A 39 7.5 10e-4.4 2.4 2.6 Crystal Structure Analysis Of The Sh2 Domain Of The Csk Homologous Kinase Chk 1PIC A 39 7.5 10e-4.4 2.7 10.3 Phosphatidylinositol 3-Kinase, P85-Alpha Subunit: C- Terminal Sh2 Domain Complexed With A Tyr751 Phosphopeptide From The Pdgf Receptor, Nmr, Minimized Mean Structure 2PFU A 38 7.0 0.0004 2.0 7.9 Nmr Strcuture Determination Of The Periplasmic Domain Of Exbd From E.Coli 2JYA A 38 6.0 0.0006 1.8 5.3 Nmr Solution Structure Of Protein Atu1810 From Agrobacterium Tumefaciens. Northeast Structural Genomics Consortium Target Atr23, Ontario Center For Structural Proteomics Target Atc1776 2FWR A 1 22 6.4 0.0010 1.3 0.0 Structure Of Archaeoglobus Fulgidis Xpb 2CHU B 2 35 5.9 0.0013 2.3 8.6 Ceue In Complex With Mecam (The top hits for just the domain VAST sees up to that point are roughly the same.) try7 search: Your VAST Search job was submitted at 07/28/2008 23:55:12(EDT). Request ID: 723694426068893421 VAST also sees a break around 124-125 in try7 (as well as one around 54-55). The full-length matches ocassionally cross those boundaries. The longest matches are PDB C D Ali. Len. SCORE P-VAL RMSD %Id Description 2F9I B 75 9.5 0.0200 1.1 6.7 Crystal Structure Of The Carboxyltransferase Subunit Of Acc From Staphylococcus Aureus 2F9I B 2 73 9.5 0.0059 0.8 6.8 Crystal Structure Of The Carboxyltransferase Subunit Of Acc From Staphylococcus Aureus 1Z8H A 44 7.1 0.0462 3.6 4.5 Crystal Structure Of Putative Lipase From The G-D-S-L Family (17135349) From Nostoc Sp. Pcc 7120 At 2.02 A Resolution 1C0G S 38 5.7 0.0263 1.8 2.6 Crystal Structure Of 1:1 Complex Between Gelsolin Segment 1 And A DictyosteliumTETRAHYMENA CHIMERA ACTIN (MUTANT 228: Q228kT229AA230YE360H) 1YNH B 3 38 5.3 0.0024 2.2 7.9 Crystal Structure Of N-Succinylarginine Dihydrolase, Astb, Bound To Substrate And Product, An Enzyme From The Arginine Catabolic Pathway Of Escherichia Coli 1VL9 A 37 5.4 0.0486 2.2 5.4 Atomic Resolution (0.97a) Structure Of The Triple Mutant (K53,56,121m) Of Bovine Pancreatic Phospholipase A2 1N28 A 35 5.4 0.0368 2.3 2.9 Crystal Structure Of The H48q Mutant Of Human Group Iia Phospholipase A2 1XAN A 3 34 5.4 0.0028 2.4 2.9 Human Glutathione Reductase In Complex With A Xanthene Inhibitor 1FEB A 4 34 5.5 0.0368 2.9 0.0 Unliganded Crithidia Fasciculata Trypanothione Reductase At 2.0 Angstrom Resolution I'll need to look at a number of these templates, to see if they offer any insight into what the fold might be for T0496. I'm too tired and bleary-eyed to do that tonight. Tue Jul 29 12:04:59 PDT 2008 Kevin Karplus I should also look at the stop few server models, to see if they offer any decent sheets that are worth polishing. The consensus metaserver MQAC selected SAM-T08-server_TS2 0.294 SAM-T08-server_TS1 0.290 These two are already included in MQAC1 and MQAU1. 3-strand sheet: 1 ^v 2 || 4 Zhang-Server_TS1 0.260 almost forms a 3-strand sheet: 1 ^v 2 ^v 3 strand 4 straight but not near sheet. RAPTOR_TS4 0.257 RAPTOR_TS3 0.255 RAPTOR_TS5 0.252 RAPTOR_TS1 0.251 almost forms a 4-strand sheet: 3 || 1 ^v 2 ^v 4 RAPTOR_TS2 0.249 3-strand sheet: 1 ^v 2 ^v 4 Zhang-Server_TS5 0.255 almost forms a 4-strand sheet: 4 || 1 ^v 2 ^v 3 Zhang-Server_TS2 0.251 almost forms a 4-strand sheet: 1 || 4 ^v 2 ^v 3 PSI_TS3 0.250 forms a 3-strand sheet: 1 ^v 2 ^v 3 BAKER-ROBETTA_TS4 0.249 Pcons_dot_net_TS3 0.249 4-strand sheet: 4 || 1 ^v 2 ^v 3 MUFOLD-MD_TS1 0.249 almost forms a 4-strand sheet: 4 || 1 ^v 2 ^v 3 There seem to be several of these forming (or almost forming) 4-strand sheets, with the 4 || 1 ^v 2 ^v 3 topology of MQAY1 and try5-opt3 being the most popular. Tue Jul 29 14:01:49 PDT 2008 Kevin Karplus Perhaps I need to try polishing the RAPTOR-like models to get another possibility. Maybe also the Zhang-Server_TS2 model. Unfortunately the "almost formed" sheets aren't close enough for undertaker to extract sheet constraints for the Zhang-Server_TS2 model, and only partial constraints are extracted for the RAPTOR models. Tue Jul 29 14:58:51 PDT 2008 Kevin Karplus I started MQAX8 metaserver run with the RAPTOR constraints, starting from servers and servers+SCWRL. I might want to try to figure out the extra constraints need to attach the other sheet and try a run with those constraints. Tue Jul 29 15:34:38 PDT 2008 Kevin Karplus Foo! MQAX8 is polishing SAM-T08-server_TS1 AGAIN. I started MQAY8 excluding the SAM servers. Tue Jul 29 21:34:34 PDT 2008 Kevin Karplus MQAY8 and MQAY9 are both optimizing Pcons_dot_net_TS3=BAKER-ROBETTA_TS4, which is 4-strand sheet: 4 || 1 ^v 2 ^v 3 not 3 || 1 ^v 2 ^v 4 So how am I going to polish the RAPTOR-like models unless I exclude all others? Tue Jul 29 22:02:31 PDT 2008 Kevin Karplus I scored all the server models with try9 and saw that the problem is that RAPTOR models have horrendous clashes. To pick one of them to polish, I'd have to turn raptor.sheets way up and soft_clashes way down. I'll try to create try10 like that and then do an MQAX10 or MQAY10 optimization (with all models or all but SAM models). Wed Jul 30 09:42:31 PDT 2008 Kevin Karplus MQAX10 is optimizing the RAPTOR_TS3-scwrl model, but it is taking forever. Looking at best models with various costfcns: try1: try6-opt3 try2-opt3 MQAX8-opt3 try7-opt3 MQAC1-opt3 MQAU1-opt3 try3-opt3 try4-opt3 try5-opt3 MQAY1-opt3 try2: try6-opt3 try2-opt3 try7-opt3 try3-opt3.gromacs0 try4-opt3 try1-opt3.repack-nonPC MQAX8-opt3.repacknonPC try5-opt3 try3: = try4: try6-opt3 try2-opt3 try7-opt3 try3-opt3.repack-nonPC try4-opt3.repack-nonPC try1-opt3.gromacs0.repack-nonPC MQAX8-opt3.gromacs0 try5-opt3 try5: try5-opt3 MQAY1-opt3.gromacs0 MQAY8-opt3 MQAY9-op3.repack-nonPC try7-opt3 try2-opt3 try3-opt3.gromacs0..repack-nonPC try1-opt3.gromacs0.repack-nonPC try6: try6-opt3 try2-opt3 try7-opt3.repack-nonPC try4-opt3.gromacs0 try3-opt3.gromacs0 MQAX8-opt3.gromacs0 try5-opt3 try1-opt3.repack-nonPC MQAY1-opt3.gromacs0 try7: try7-opt3 try6-opt3 try3-opt3.gromacs0 try2-opt3.gromacs0 try5-opt3.groamcs0 MQAY1-opt3.gromacs0 MQAY8-opt3.gromacs0 MQAY9-opt3.repack-nonOC try8: try5-opt3 MQAY1-opt3.gromacs0 try7-opt3 try6-opt3 MQAY8-opt3 try2-opt3 MQAY9-opt3 try3-opt3.gromacs0 try4-opt3.gromacs0 try9: try5-opt3 MQAY1-opt3.gromacs0 MQAY8-opt3 MQAY9-opt3 try7-opt3 try6-opt3 try2-opt3 try3-opt3.gromacs0 try4-opt3.gromacs0 try10: MQAX10-opt2 try5-opt3 MQAY1-opt3 try6-opt2 try4-opt3 try7-opt3 try2-opt3.repack-nonPC try3-opt3 rosetta:try5 MQAY9 MQAY1 MQAY8 try7 try4 try6 try2 try3 try1 Wed Jul 30 11:30:24 PDT 2008 Kevin Karplus 6 models in best-models have the same 3-strand sheet 1 ^v 2 || 4: try7 try6 try3 try4 MQAX8 MQAC1 Also similar to each other with a 4-strand sheet 4 || 1 ^v 2 ^v 3 MQAY8 MQAY9 Other models are try5 4-strand sheet: 4 || 1 ^v 2 ^v 3 MQAX10 3 ||? 1 ^v 2 ^v 4 (strand 3 not fully formed and has bad breaks) The try5 and MQAY8 sheets are slightly different: try5: SheetConstraint I126 V128 K124 I126 hbond V128 1 SheetConstraint D96 N102 R56 A62 hbond N102 1 SheetConstraint F70 F73 T80 V77 hbond H71 1 SheetConstraint V59 C64 F73 T68 hbond M60 1 MQAY8/MQAY9: SheetConstraint I126 V128 K124 I126 hbond V128 1 SheetConstraint E95 F103 R56 C64 hbond F103 1 Phase difference SheetConstraint H71 K74 T80 V77 hbond F72 1 Phase difference SheetConstraint G57 C64 K74 D67 hbond L58 1 Phase difference I like the sep=3 turns better than the sep=4 turns, so I should try to polish MQAY8 and MQAY9. Wed Jul 30 12:30:31 PDT 2008 Kevin Karplus try11 started to polish MQAY8 and MQAY9. try11 scores quite well, and Rosetta likes it almost as well as try5. I should probably do one more polishing run with clashes and sidechains turned up, to see if I can get it to pull ahead of try5. Wed Jul 30 15:32:13 PDT 2008 Kevin Karplus OK, try12-opt3 is the new best score with the try12 costfcn and try12-opt3.gromacs0.repack-nonPC is rosetta's favorite. I'll submit that as my model 3, the less-good 4-strand sheet (try5) as model 5, and include 3 instances of the 3-strand sheet for my other models. I gave up on the RAPTOR-like models. Wed Jul 30 15:51:39 PDT 2008 Kevin Karplus Submitted with comment: The SAM HMMs came up with no hits for T0496--the best Evalue was 41.6 for 2f9iB. I ended up selecting models by how much beta-sheet they formed: how compatible they were with secondary-structure prediction. There were two lineages of models: those that came from SAM alignments and SAM servers, which had a 3-strand sheet 1 ^v 2 || 4 and those that came from Robetta with a 4-strand sheet: 4 || 1 ^v 2 ^v 3 I also considered the almost-a-sheet models created by RAPTOR with 3 ||? 1 ^v 2 ^v 4, but I could not get consistent models with that topology in the time available, so did not submit any. model 1 T0496.try7-opt3.pdb # < chimera-try6-try3 chimera-try6-try3: M1-N119 from T0496.try6-opt3.pdb P120-L178 from T0496.try3-opt3.repack-nonPC.pdb try6-opt3 < try2-opt3 < try1-opt3 < align(1ryp2+2f9iB+2chcA+1ixsA+2i9cA) try3-opt3 < align(1neuA+2f9iB+1ibrB+2i1mA+3bqoA+1ixsA+1q2hA) 2 T0496.try6-opt3.pdb # < try2-opt3 < try1-opt3 < align(1ryp2+2f9iB+2chcA+1ixsA+2i9cA) 3 T0496.try12-opt3.gromacs0.repack-nonPC.pdb # < MQAY8-opt3.repack-nonPC < Pcons_dot_net_TS3=BAKER-ROBETTA_TS4 # best rosetta score 4 T0496.try4-opt3.gromacs0.pdb # < align(1egaA+1p0zA+1ixsA+2f9iB+1q2hA+2i9cA+1ht6A) 5 T0496.try5-opt3.gromacs0.repack-nonPC.pdb #