Fri Aug 16 16:04:57 PDT 2002 t0194 18 August Rachel Karchin No great template hits from the combined models, but the single-track model is picking up a weakly good hit (which is in our template library). 2pgd gets E-value of 0.757 2pgd has two SCOP domains a.100.1.1 c-terminal domain-like 5-phosphogluconate dehydrogenase all-alpha, two structural repeats, 5 helices each c.2.1.6 n-terminal domain 6-phosphogluconate dehydrogenase-like a/b beta-alpha-beta, 8-stranded beta sheet There are eight putative beta strands in the STR secondary structure prediction for T0194 as well. T0194 has 237 residues, so it could be multi-domain, however the secondary structure looks like a/b, so my guess is that the hit is to c.2.1.6, the n-terminal domain (residues 1-177 of 2pgd) CAFASP link to this target is broken (pointing to T0193!) T0194.try1-opt beta sheets are blown apart. loosely packed structure with holes. 2pgd N-terminal domain looks like a reasonable template. I'll generate alignments to 2pgd and add to the mix. Now starting undertaker try2 with 2pgd alignments included . .. 20 August 2002 Rachel Karchin Rough definition of secondary structures: E17-V28 H1 D32-I39 E1 S44-D56 H2 K60-D65 E2 R67-T86 H3 R91-K103 H4 A107-V111 E3 T114-A119 H5 L123-S133 H6 L139-A141 E4 ?? L149-D153 E5 H159-L173 H7 R182-T188 E6 E192-L206 H8 N211-S218 E7 V220T225 E8 S229 Q235 E9 try2-opt.scwrl is new best scorer. Looks much better than try1-opt.scwrl. Parallel sheet forming with these strands: E1 E6 E3 Fri Aug 30 13:43:37 PDT 2002 Rachel Karchin I'm going to try to lock in the secondary structure features I like in try2-opt.scwrl with hbond constraints and then do another run, reading this in as a conformation. ideas -- do a str-heavy alignment do global alignment to n-terminal domain (of 2pgd only) chop off c-term from stride-mixed seq and stride-mixed.str do global alignment of 2-track target model aa-str to these Wed Sep 4 15:31:43 PDT 2002 Rachel Karchin best scorers: T0194.try2-opt-scwrl.pdb This one has a four-stranded sheet forming. Also two hairpins, a possible hairpin that's curling up The sheet could pack against two helices S44-D56 and E17-V28. The region with hairpins is more problematic as the hairpins are just sticking out into solvent? Hydrophobic residues on these hairpins are pointing out, so it looks like orientation is all wrong in this region. T0194.try3-opt-scwrl.pdb building str-heavy alignments . . . do global alignment to n-terminal domain (of 2pgd only) This is residues A1-G177. Created 2pgd-nterm.dssp.2d 2pgd-nterm.dssp.str 2pgd-nterm.seq created various alignments to single-track and stride and str two-track models. try4 -- just work with all the new alignments, and continue with straightness constraints Sun Sep 8 09:18:21 PDT 2002 Rachel Karchin best scoring is still T0194.try2-opt-scwrl.pdb, although the third-ranked decoy try4-opt-scwrl has formed a better-looking three-stranded parallel sheet for L33-I39 R182-T187 A106-V111 Also has two pairs of hairpins formed at S141-N146 F136-L138 and V220-T225 S229-Q235 (C-terminus) The C-terminal pair could use some constraints: -- to straighten P219-I221 -- and hold the hairpin together better That being done, this hairpin could be part of the three-stranded sheet. But before that step, I see there is a predicted strand F61-D65 that is lying right next to the three-stranded sheet (adjacent to A106-V111) and is a likely fourth strand in the sheet. The C-terminal hairpin could form the fifth and sixth strands . . . To do: Alignment for: L33-I39 R182-T187 A106-V111 F61-D65 try2-opt-scwrl has this four-stranded parallel sheet and it has a fifth straight piece V8-T15 that looks like it could be a fifth strand in the sheet: It also has formed somewhat broken versions of three hairpins which are close enough together in space that they could form a mixed or anti-parallel six-stranded sheet. Todo: align the four-(five?)-stranded sheet and the putative six-stranded sheet. Come up with constraints. Use try2-opt-scwrl and try3-opt-scwrl as starting conformations for try5. Mon Sep 9 21:04:04 PDT 2002 Rachel Karchin five-stranded parallel sheet gVKFSTSQTt ydLITGIPIdke kRGAFLTTVfd taAVLFVtd askFHVLDq Hmm. I don't have any idea where the hairpins should go in the structure. try4 will just have constraints for the above sheet. 10 Sept 2002 10:07 Kevin Karplus Since try4-opt has a pretty good sheet, with 3 solid strands and another strand almost attached, I'm going to submit it to VAST and see what templates it resembles. Job VS31418 password casp5t0194 Check the results at http://www.ncbi.nlm.nih.gov/cgi-bin/Structure/vastresults?result=Here in an hour or so. Tue Sep 10 16:50:46 PDT 2002 Rachel Karchin VAST hits: whole chain: RMSD NRES %ID 1e32A 1.6 68 8.8 1d2hA 0.6 76 7.9 domain1 1e32A 1.8 69 8.7 1d2hA 1.0 83 8.4 1d2hA 0.7 76 7.9 1g6oA 1.4 46 15.2 1iqpA 1.9 53 5.7 1jbkA 1.6 43 11.6 1d2mA 1.6 41 0.0 Cool! A hit against a protein with 0% sequence identity. We didn't fish out any of these structures with our HMMs. I'm going to add alignments for any of them that are in the models.97 library. grunt 108> ../scripts-human/in-models97 1e32A 1e32A models found grunt 110> ../scripts-human/in-models97 1g6oA 1g6oA models found grunt 112> ../scripts-human/in-models97 1jbkA 1jbkA models found grunt 113> ../scripts-human/in-models97 1d2mA 1d2mA models found Also want to add the VAST alignments . . . Wed Sep 11 12:03:14 PDT 2002 Rachel Karchin try6 started -- uses same constraints as try5 but incorporates alignments from the templates found by VAST. From karplus@bray.cse.ucsc.edu Wed Sep 11 13:27:55 2002 Date: Wed, 11 Sep 2002 13:27:52 -0700 From: Kevin Karplus To: rachelk@soe.ucsc.edu CC: karplus@soe.ucsc.edu Subject: t0194 Because you started try6 with the already polished try5, the new alignments aren't going to do much. I'll start a similar run (try7) which does NOT have the try5-opt alignment as a starting point. ---------------------------------------- 14:40 Kevin Karplus Because try6 and try7 are taking a long time, I'm going to start a short polishing run try8 (with the current try5-constraints scoring function) from the best current decoys: try6.1.60 try6.0.60 try5-opt-scwrl. I noticed that the C-terminal antiparallel sheet is starting to form, but has no constraints to aid it. I've added some in try9.constraints, and will I will do another polishing run with those constraints. With the try9 constraints, try2-opt-scwrl scores best, but I'm going to start with the same decoys as for try8, since they are doing best on the parallel sheet. ------------------------------------------ Sept 9 Jenny Draper & Johnathan Casper 3:30pm T0194.try7+T0194-1d2nA-vit-adpstyle5.pw.a2m.gz:1d2nA.0.20.pdb has a nice parallel sheet, but all it's helices are broken off... 1d2nA has a few VAST matches, so we should be getting a better alignment to it soon... (the alignment this decoy is from is one of the HMM alignments Rachel generated last night). The strand ordering is (strand definitions approximate & not lined up) 3> K60 - L64 4> A107 - V111 7> G183 - T188 2> L33 - P38 It also looks quite accomodating for the 3-strand sheet at the C-Term end... it has the last haipin and & third strand straight... It could even have a 5-6-8-9-10 or 8-9-10-5-6 all AP sheet... From karplus@bray.cse.ucsc.edu Wed Sep 11 16:50:27 2002 Date: Wed, 11 Sep 2002 16:49:27 -0700 From: Kevin Karplus To: karplus@soe.ucsc.edu, rachelk@soe.ucsc.edu, weber@soe.ucsc.edu, learithe@cats.ucsc.edu, yael@biology.ucsc.edu, baertsch@soe.ucsc.edu, rph@soe.ucsc.edu, afyfe@soe.ucsc.edu, jcasper@soe.ucsc.edu, oscarhur@soe.ucsc.edu Subject: t0193 and t0194 For t0194, the best so far seems to be try9-opt. Any ideas where to tuck strands H137-F142 and L149-D153? ------------------------------------------------- Sept 9 Jenny Draper 5:30 Running a try10 off of the alignments to 1d2nA (including the VAST alignments), with no constraints.:w Date: Wed, 11 Sep 2002 17:37:13 -0700 (PDT) From: Jenny Draper To: Kevin Karplus cc: rachelk@soe.ucsc.edu, weber@soe.ucsc.edu, yael@biology.ucsc.edu, baertsch@soe.ucsc.edu, rph@soe.ucsc.edu, afyfe@soe.ucsc.edu, jcasper@soe.ucsc.edu, oscarhur@soe.ucsc.edu Subject: Re: t0193 and t0194 I think strands H137-F142 and L149-D153 should be a hairpin, probably tucked into the the sheet with 8-9-10 in antiparallel. As to which side, I have no clear idea. ---------------------------------------------------------------------- 17:54 Kevin Karplus I'll start try11 with hairpin // 134> lkfHLLISFlg // 156< rakDFFQLKgne added to the try9 constraints. I don't know where it should end up either, so let's just try to form the hairpin. ---------------------------------------------------------------- 11 Sept 2002 Jonathan Casper 6:22pm Date: Wed, 11 Sep 2002 18:22:02 -0700 (PDT) From: Jonathan Casper To: Kevin Karplus Subject: Re: t0193 and t0194 On Wed, 11 Sep 2002, Kevin Karplus wrote: > > For t0194, the best so far seems to be try9-opt. > Any ideas where to tuck strands H137-F142 and L149-D153? > It looks like the final strand is trying to be in-between the other two (the final three-strand anti-parallel constraints in try9.constraints). Could the other two strands bracket that three-strand sheet? The H137 strand could form up parallel to the N211 strand, and the L149 strand could run parallel to the 219 strand (if the final 232 strand goes in the middle), or anti-parallel to the 232 strand. Jonathan ------------------------------------------------------------------- Jenny Draper 6:45pm Just started try12, doing the same thing as try10 but with constraints added from the T0194.try7+T0194-1d2nA-vit-adpstyle5.pw.a2m.gz:1d2nA.0.20.pdb sheet, and the antiparallel sheet in the runs Kevin is working on. Also started try13, which should be trying to improve the try7-1d2nA alignment. I'm going out to dinner; I should be back to check on the runs in awhile... ...forgot to fix a few errors in my CB definitions of try12.constraints. Shouldn't be a problem, since they're redundant; should be fixed for scoring, but don't have time at the moment... 19:00 Kevin Karplus Looking at try11-opt, I think I see Jonathan's point aobut the two strands. Let me try to put together a try14 constraints with // 134> lkfHLLISFlg // 208> hsiNIEQSLLsp // 156< rakDFFQLKgne // 226< gTHTCIV // 227> lnSYVIVQge 20:38 I was not impressed with try14-opt, so I'm trying again on try15, this time putting the hairpin on one side of the 3-strand sheet: // 208> hsiNIEQSLLsp // 226< gTHTCIV // 227> lnSYVIVQge // 156< rakDFFQLKg // 134> lkfHLLISFlgen Right now, my favorite is try11-opt. best with try15.constraints: try11-opt try14-opt try13.0.100 try9-opt try14.constraints: try14-opt try11-opt try13.0.100 try5.10.60 try12.constraints: try13.0.100 try13.1.100 try12-a10.0.100 try11-opt try11.constraints: try11-opt try14-opt try13.0.100 try9-opt try9.constraints: try11-opt try9-opt try13.0.100 try2-opt-scwrl try5-constraints: try11-opt try8-opt try9-opt try6.1.60 try3-constraints: try2-opt-scwrl try5.15.60 try5.5.60 try3-opt-scwrl no constraints: try8-opt try11-opt try9-opt try6.1.60 try13.0.100 has a hairpin starting the 3-strand-sheet with Hbonds between H224 and Y230 and between C222 and I232, better than the hairpin in try11-opt, but the other strand of the sheet is much worse. Until the try15 run is finished, my favorite is try11-opt. Current proposal: model 1: try11-opt model 2: try14-opt model 3: try13.0.100 model 4: try8-opt try15-opt finished, and may be slightly better than try11-opt, scoring better than it with try15 constraints and with no constraints, but it did not do what I was trying for. I suspect that the helix E70-L84 should have its hydrophobic face packed against the large sheet, rather than sticking the helix out into space, but I don't have the time or the patience to try to do anything about it in the remaining hour or two. I'll submit model 1: try15-opt model 2: try11-opt model 3: try13.0.100 model 4: try8-opt model 5: try14-opt 11 Sept 2002 Jenny Draper 10pm looks good to me. i could go either way with switching models 2 & 3, especially since try13 is from alignments, but I'm OK with this. :) From: Rachel Karchin Date: Wed, 11 Sep 2002 22:32:44 -0700 To: Kevin Karplus Subject: Re: t10193 done, t0194 waiting for one more run to finish T0193 model 1: try15-opt model 2: try11-opt model 3: try13.0.100 model 4: try8-opt model 5: try14-opt I don't see much difference between try15-opt and try11-opt. try13.0.100 is more of an alternative -- not that compact, but the hairpins and three-stranded sheet look good. I'd vote for switching the order of try11-opt and try13.0.100. ------------------------------------------------------------------------------ 12 Sept 2002 Kevin Karplus 02:13 Rachel, I sent the models off before I got your message, and went to bed. It doesn't really matter though, since there are really only two evaluations done---evaluation of model 1 and evaluation of all 5 models. Had you wanted a different model as model 1 or a different model put in the top 5, it would have made more of a difference. From karplus@bray.cse.ucsc.edu Thu Sep 12 02:16:23 2002 Date: Thu, 12 Sep 2002 02:16:20 -0700 From: Kevin Karplus To: rachelk@soe.ucsc.edu CC: karplus@soe.ucsc.edu In-reply-to: <15744.9980.320932.349062@grunt.cse.ucsc.edu> (message from Rachel Karchin on Wed, 11 Sep 2002 22:32:44 -0700) Subject: Re: t10193 done, t0194 waiting for one more run to finish Rachel, I sent the models off before I got your message, and went to bed. So the order was the one I put in the README file. It doesn't really matter though, since there are really only two evaluations done---evaluation of model 1 and evaluation of all 5 models. Had you wanted a different model as model 1 or a different model put in the top 5, it would have made more of a difference. Kevin