Wed Aug 4 15:17:21 PDT 2004 T0263 DUE 27 Aug 2004 Fri Aug 6 14:24:24 PDT 2004 Kevin Karplus Because of a bug I introduced to Make.main, I had to remake this prediction. This looks like a fold-recognition target. The best hits with t2k are 1iujA, 1lq9A, .... it looks like a d.58.* fold. T04 concurs, though putting 1lq9A first. I'll make extra_alignments so that the next few t04 hits are included in the list of possible templates, though these are likely to be wrong. Fri Aug 6 16:53:09 PDT 2004 Kevin Karplus try1 looks pretty good. I think I want to add an antiparallel sheet constraints for T3-A10 antiparallel D87-E80. I don't think I want to specify hbonding yet, since some of the strands may be flipped over still. The helices look well-placed and oriented, so this may not require much polishing. (The robetta models seem to make more complete sheets, but the helices aren't good.) From karplus@soe.ucsc.edu Mon Aug 16 22:30:04 2004 Date: Mon, 16 Aug 2004 22:30:03 -0700 From: Kevin Karplus To: martina@soe.ucsc.edu Cc: karplus@soe.ucsc.edu In-reply-to: (message from Martina Koeva on Mon, 16 Aug 2004 22:20:04 -0700 (PDT)) Subject: Re: T0263 You wrote: > T0263 is one of the other targets that I am working on due next week. I > have been looking at the templates. It seems to me from going through the > literature that our top hits are dimers. We are forming a pretty good > monomer right now. The strands are arranged in a sheet with helices > packing on one side only. The templates suggest that the dimer interface > would be between the sheets (which are predicted to be buried on the side > not facing the sheet). > > My question is the following: If I believe that the structure should be a > dimer, are we usually trying to build a dimer so that we can fix our > monomer submissions,or do we actually submit a dimer? Would you advise me > in this case to try and build a dimer? We can't submit dimers, but we can build monomers that are consistent with dimerization. In such cases it is generally best to build the dimer by superposition on a dimeric template, then look for pieces we can improve to avoid clashes or to make better contact with the dimerization partner. Both Sol and I have built a number of dimers, and it is fairly straightforward to do. Kevin ------------------------------------------------------------------------ Date: Tue, 17 Aug 2004 19:37:50 -0700 (PDT) From: Martina Koeva To: Kevin Karplus Subject: T0263 again Hi Kevin, I am back to my dimer question: Our top template, as well as two other templates that I like(1lq9A, 1iujB, 1rjjA) are all homodimers. The monomers for each one consist of 4-stranded antiparallel sheets, where the last fourth strand is a "swapped" strand and belongs to the other monomer. I followed the instructions in the main README file and right now I have a dimer structure, which does not though have those strands swapped in T0263, because the latest monomer model that we have does not extend like that. It looks pretty happy to form its whole sheet from its own strands The segment that I am interested in using for the swapping is very weakly predicted to be a helix and is just flopping around right now on the side of the sheet. I went back to T0229, in which you and Sol were working on scaffolding constraints for that dimer. Do you have any advice for me in regard to how to go about that in T0263? Thank you! Martina ----------------------------------------------------------------------- Date: Tue, 17 Aug 2004 19:47:12 -0700 (PDT) From: Martina Koeva To: Kevin Karplus Cc: Martina Koeva Subject: Re: T0263 again Hmmm, I see in the T0229 README file that Sol used this "3-point method" for the scaffolding constraints and I remember you mentioned that it was successful. I guess I am going to go ahead and try that. Martina ------------------------------------------------------------------------ Date: Tue, 17 Aug 2004 20:53:58 -0700 From: Kevin Karplus To: martina@soe.ucsc.edu Cc: karplus@soe.ucsc.edu Subject: Re: T0263 again I don't have any advice on T0263 at the moment. If you are getting an unswapped monomer, we should probably submit that as one model, in addition to any dimeric models you get. ----------------------------------------------------------------------- Tue Aug 17 23:12:57 PDT 2004 Martina Koeva Here is a summary of all the information that I have so far for T0263. I decided to start from our top hit with t04 and t2k - 1lq9A, as well as another template 1iujA. I also ran VAST on try2 of the structure (I have lost the VAST ID, which means that I would have to resubmit the structure so that we have it for the record). VAST found a couple of hits, including 1lq9A, 1iujB, 1rjjA. The common feature of all these structures is their ferrodoxin-like fold. In particular, they are all homodimers. Each monomer consists of 3 helices and 4 strands, oriented in an antiparallel manner (ferrodoxin-like split beta-alpha-beta fold with an antiparallel beta sheet). The dimer is stabilized through C-terminal strand swapping. The 4th beta strand at the C-terminus contributes to the beta sheet of the other monomer by laying antiparallel to the second strand (in ascending residue number). In the templates (at least the literature shows it's true for 1lq9A) the swapped strand participates in the formation of the active site. It contributes through I110 (in 1lq9A; in T0263 that corresponds to L100). Here is: 1.) a link to the paper for 1lq9A: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&dopt=Abstract&list_uids=12514126 2.) a link to the paper for 1iujA/B http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&dopt=Abstract&list_uids=15103643 3.) a link to the paper for 1rjjA http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=pubmed&dopt=Abstract&list_uids=15213437 If one looks at the paper for 1iujA, they have included an alignment of all these three structures that share not only the fold, but also the swapped C-terminal strand. I built a homodmier from try2-opt2 and named it dimer-1lq9-try2.pdb. This was done following the procedure, described in the main README. The alignment to 1lq9 has been included in the 1lq9A subdirectory under: 1lq9.2mer-a2m. The script used to build the dimer was named make-dimer-try2.under. The current dimer though does not swap strands between the two monomers. The part of the strand (from the alignments) that I expect to see the swapping happen in, corresponds to the segment from around residue 90 to residue 100. I looked at our str2 and ehl2 predictions and we have very weak helical predictions for that segment. In try2-opt2 that segment looks like is flopping around on the side(or above) of the sheet. This lead me to look at our t04 and t2k bys predictions (or the t04 and t2k bys.constraints files generated), as I have found before that I can pick up extra constraints that fit what I am looking for. The bys.constraints files agreed (more or less) on the following: # from t04.bys.constraints StrandConstraint E90 P91 0.684 StrandConstraint A98 A101 0.635 I have included those in try3.costfcn. In order to get though this proposed swapped strand, I have decided to go ahead and use scaffolding constraints. The approach in choosing those scaffolding constraints follows, Sol's approach described in the README file for T0229: I chose 3 points (residues) in the swapped strand) and 3 conserved residues: # residues in swapped strand: # T0263 1lq9A # ------------- # L100 I110 # A98 G108 # Q97 R107 # # conserved residues off the swapped strand # T0263 1lq9A # ------------ # Y54 A64 # W41 Y51 # F36 F46 end of very well-conserved beta turn I also used Rasmol to measure the appropriate distances in 1lq9A. # distances measures in 1lq9A # # Distance A64B.CA-I110B.CA: 19.136 # Distance A64B.CA-R107B.CA: 15.542 # Distance A64B.CA-G108.CA: 16.874 # # Distance Y51B.CA-I110B.CA: 22.007 # Distance Y51B.CA-R107B.CA: 21.854 # Distance Y51B.CA-G108.CA: 22.160 # # Distance F46B.CA-I110B.CA: 23.717 # Distance F46B.CA-R107B.CA: 18.991 # Distance F46B.CA-G108.CA: 20.682 These distances translated into the following scaffolding constraints, introduced in try3.costfcn: Constraint Y54.CA L100.CA 17.136 19.136 21.136 1.8 Constraint Y54.CA Q97.CA 13.342 15.342 17.342 1.8 Constraint Y54.CA A98.CA 14.874 16.874 18.874 1.8 Constraint W41.CA L100.CA 20.007 22.007 24.007 1.8 Constraint W41.CA Q97.CA 19.854 21.854 23.854 1.8 Constraint W41.CA A98.CA 20.160 22.160 24.160 1.8 Constraint F36.CA L100.CA 21.717 23.717 25.717 1.8 Constraint F36.CA Q97.CA 16.991 18.991 20.991 1.8 Constraint F36.CA A98.CA 18.682 20.682 22.682 1.8 What kind of weight does one give these constraints? I have setup try3 and it is running on hoot (hope this works!). Once it is done, I will attempt to build another dimer, using try3-opt2. I have also updated the cost function for try3, such that the dry weights are much higher now as compared to try2.costfcn. I have also turned down the weight on sidechain (from 6 back to 1). Tue Aug 17 23:59:25 PDT 2004 Martina Koeva Hmm, maybe I chose the wrong residues to do the scaffolding...In try3-opt1 the helix is still a helix and is tucked in nicely behind the sheet, instead of in front of it. :( I will have to wait to see what try3-opt2 does. Wed Aug 18 02:49:25 PDT 2004 Martina Koeva So try3-opt2 confirmed what I saw in try3-opt1. I am inclined to think that the residues that I chose were not very appropriate, since they were pointing in the direction opposite of the dimer interface. For constraints in try4 I have selected only residues that face the dimer interface: # residues in swapped strand: # T0263 1lq9A # ------------- # A101 L111 # K99 A109 # N96 P96 # # conserved residues off the swapped strand # T0263 1lq9A # ------------ # G51 V61 # I53 Y63 # L4 V14 # # # distances measures in 1lq9A # # Distance VAL61B.CA-LEU111B.CA: 22.396 # Distance VAL61B.CA-ALA109B.CA: 20.370 # Distance VAL61B.CA-PRO106B.CA: 17.469 # Distance TYR63B.CA-LEU111B.CA: 18.217 # Distance TYR63B.CA-ALA109B.CA: 15.661 # Distance TYR63B.CA-PRO106B.CA: 12.051 # Distance VAL14B.CA-LEU111B.CA: 20.646 # Distance VAL14B.CA-ALA109B.CA: 16.956 # Distance VAL14B.CA-PRO106B.CA: 11.965 Here is my new set of constraints, as applied to T0263: Constraint G51.CA A101.CA 20.396 22.396 24.396 1.8 Constraint G51.CA K99.CA 18.370 20.370 22.370 1.8 Constraint G51.CA N96.CA 15.469 17.469 19.469 1.8 Constraint I53.CA A101.CA 16.217 18.217 20.217 1.8 Constraint I53.CA K99.CA 13.661 15.661 17.661 1.8 Constraint I53.CA N96.CA 10.051 12.051 14.051 1.8 Constraint L4.CA A101.CA 18.646 20.646 22.646 1.8 Constraint L4.CA K99.CA 14.956 16.956 18.956 1.8 Constraint L4.CA N96.CA 9.965 11.965 13.965 1.8 I am including them in try4.costfcn. I will also need to increease the weight on the strand constraints for that last segment. Wed Aug 18 12:58:04 PDT 2004 Martina Koeva I looked at the results from try4 and try4-opt2 looks more promising. The C-terminal segment is somewhat more extended than before, even though it still comes down very close to its own monomer. This makes me think I should include a couple of extra constraints for the residues on that swapped strand. The template 1lq9 has a very well-conserved loop with Pro44 and Gly45, which corresponds to Pro34 and Gly35 in T0263. I will introduce these extra 6 constraints for try5, as well as raise the strand constraint weights. I really need that segment to extend like a strand. # distances measured in 1lq9 # Distance PRO44B.CA-PRO106B.CA: 19.415 # Distance PRO44B.CA-ALA109B.CA: 25.087 # Distance PRO44B.CA-LEU111B.CA: 27.937 # Distance GLY45B.CA-PRO106B.CA: 15.636 # Distance GLY45B.CA-ALA109B.CA: 21.418 # Distance GLY45B.CA-LEU111B.CA: 24.572 Constraint P34.CA A101.CA 25.937 27.937 29.937 1.8 Constraint P34.CA K99.CA 23.087 25.087 27.087 1.8 Constraint P34.CA N96.CA 17.415 19.415 21.415 1.8 Constraint G35.CA A101.CA 22.572 24.572 26.472 1.8 Constraint G35.CA K99.CA 19.418 21.418 23.418 1.8 Constraint G35.CA N96.CA 13.636 15.636 17.636 1.8 Wed Aug 18 15:34:43 PDT 2004 Martina Koeva Try5 is done and at least now is not folded into a helix, even though it is still flopping around (but at least in the right direction). I used the try5-opt2 model to create a dimer (try3) version, which has been saved under dimer-1lq9-try3.pdb. The new dimer version looks better - it is starting to extend the C-terminal segment in an antiparallel fashion next to the appropriate strand on the opposite monomer. Wed Aug 18 16:16:37 PDT 2004 Martina Koeva All right, I have a new set of constraints (I feel like I am killing this structure with the number of constraints I am putting in). I selected these new residues mainly to represent diverse points on the main sheet, as well as points that stick out toward the dimer interface side. # Distance VAL102B.CA-PRO106B.CA: 12.656 # Distance VAL102B.CA-ALA109B.CA: 18.680 # Distance VAL102B.CA-LEU111B.CA: 23.377 # Distance HIS52B.CA-PRO106B.CA: 19.212 # Distance HIS52B.CA-ALA109B.CA: 20.453 # Distance HIS52B.CA-LEU111B.CA: 21.033 # Distance LEU47B.CA-PRO106B.CA: 13.702 # Distance LEU47B.CA-ALA109B.CA: 17.869 # Distance LEU47B.CA-LEU111B.CA: 20.004 Constraint L37.CA A101.CA 18.004 20.004 22.004 1.8 Constraint L37.CA K99.CA 15.869 17.869 19.869 1.8 Constraint L37.CA N96.CA 11.702 13.702 15.702 1.8 Constraint T42.CA A101.CA 19.033 21.033 23.033 1.8 Constraint T42.CA K99.CA 18.453 20.453 22.453 1.8 Constraint T42.CA N96.CA 17.212 19.212 21.212 1.8 Constraint L92.CA A101.CA 21.377 23.377 25.377 1.8 Constraint L92.CA K99.CA 16.680 18.680 20.680 1.8 Constraint L92.CA N96.CA 10.656 12.656 14.656 1.8 I am a bit worried about the number of constraints that I am putting into the cost function. On the other hand, as try5-opt2 shows, the structure itself is not completely overt to adopting the conformation that I am forcing it in. I wonder whether that is a good idea. Thu Aug 19 13:46:17 PDT 2004 Martina Koeva I have started try7 keeping the cost function the same, but starting from existing models. Thu Aug 19 16:36:08 PDT 2004 Martina Koeva I do not like the hydrgogen bonding that does not allow the C-terminal segment to extend fully. I will introduce one final constraint to extend that segment # Distance THR105B.CA-ALA101B.CA: 13.413 Constraint P91.CA I95.CA 11.413 13.413 15.413 1.8 I have setup try8 to run from existing models again and have included this final constraint. Fri Aug 20 16:42:29 PDT 2004 Martina Koeva I started try9 with increased distance constraints (from 1.8 to 3) on heehaw. Try9 though crashed for an unknown reason and have copied the .costfcn and .under file into a try10 version and have restarted the try on heehaw again. Sat Aug 21 11:13:57 PDT 2004 Kevin Karplus I'm trying to reproduce Martina's assertion failure running under gdb. My first attempt did not work, because I forgot to save the debug9.under file with the seed set. THe second try has the seed set and appears to be doing the same computations (the best score lines still match at generation 54). Sat Aug 21 11:22:30 PDT 2004 Kevin Karplus They differ at generation 4 of iteration 2, so something different has happened. Note that this difference occurs long before the assertion failure in try9, so the error is probably no where near the place the assertion failure happens. Sure enough, this run sails right on past the point where try9 had the assertion failure. It looks like I will have to do a run under valgrind to find the uninitialized variable problem (assuming that is what causes the non-deterministic behavior). Sat Aug 21 12:12:49 PDT 2004 Kevin Karplus Ahh--I do get an assertion failure under gdb! I found a divide-by-zero error that was the proximate cause of the assertion failure, but I haven't figured out what is causing the non-deterministic behavior. Sun Aug 22 19:25:03 PDT 2004 Martina Koeva Try11 finished and neither try9, nor try11 models seem to be improving the structure a bit. I have used the try7-opt2 model to create a new version of the dimer (dimer-1lq9-try4.pdb), using the script make-dimer-try4.under. This version of the dimer is pretty much the same as the dimer-1lq9-try3.pdb,which used try5-opt2 as the monomer model for this homodimer. Mon Aug 23 18:19:08 PDT 2004 Martina Koeva I have setup a dimer directory, in which we put together the two monomers in the dimer as a single chain. I have started a setup run, using the sheet constraints picked up from dimer1.sheets on the dimer-as-monomer.pdb file (I ran a extract-dimer-sheets.under). I have also added additional sheet constraints to pair the C-terminal strand in the original monomer to the edge strand on the other side. The alignment of those strands that I am looking for is: 37: LWKVWT :42 (edge strand on the opposite monomer) | 202:ALKAQN:197 (C-terminal strand on the original monomer) Tue Aug 24 12:44:29 PDT 2004 Sol Katzman After talking to Martina, I decided to give a try at the scaffolding constraints. Same idea as before, and I am starting from try8, so I will call mine try9s. Note: to avoid possible collisions, I will do the repack and score-all runs separately, only after the try9s-opt2.pdb.gz run is done. Also, you may want to be careful about using read-pdb.under from now on. # 1lq9A T0263 # ---- ----- # these are at the end of a sheet on 3 strands -- the anchor # F10 T3 # W65 F56 # F45 F36 # # these are away from the end of the sheet, with one in a different plane # T15 H8 # A77 E67 # these are based on hbonds to determine strand location # 11hbond103 <-> 4hbond88, therefore P105 <-> E90 # we want them to dimerize to the other monomer # P105 E90 # R106 P91 # G107 L92 # A108 S93 # I109 Q94 # L110 I95 # sol attempt at scaffolding constraints from 1lq9A # Distance PRO105A.CA-PHE10A.CA: 7.136 # Distance PRO105A.CA-TRP65A.CA: 10.819 # Distance PRO105A.CA-PHE45A.CA: 15.516 # # Distance ARG106A.CA-PHE10A.CA: 10.379 # Distance ARG106A.CA-TRP65A.CA: 14.344 # Distance ARG106A.CA-PHE45A.CA: 19.045 # # Distance GLY107A.CA-PHE10A.CA: 13.224 # Distance GLY107A.CA-TRP65A.CA: 16.403 # Distance GLY107A.CA-PHE45A.CA: 20.748 # # Distance ALA108A.CA-PHE10A.CA: 14.596 # Distance ALA108A.CA-TRP65A.CA: 16.826 # Distance ALA108A.CA-PHE45A.CA: 20.668 # # Distance ILE109A.CA-PHE10A.CA: 18.116 # Distance ILE109A.CA-TRP65A.CA: 19.903 # Distance ILE109A.CA-PHE45A.CA: 23.363 # # Distance LEU110A.CA-PHE10A.CA: 19.453 # Distance LEU110A.CA-TRP65A.CA: 20.276 # Distance LEU110A.CA-PHE45A.CA: 23.059 # Distance LEU110A.CA-THR15A.CA: 25.320 # Distance LEU110A.CA-ALA77A.CA: 34.694 # # Distance ILE109A.CA-THR15A.CA: 24.322 # Distance ILE109A.CA-ALA77A.CA: 33.922 # # Distance ALA108A.CA-THR15A.CA: 22.407 # Distance ALA108A.CA-ALA77A.CA: 30.612 # # Distance GLY107A.CA-THR15A.CA: 22.770 # Distance GLY107A.CA-ALA77A.CA: 29.428 # # Distance ARG106A.CA-THR15A.CA: 21.362 # Distance ARG106A.CA-ALA77A.CA: 26.368 Constraint E90.CA T3.CA 5.136 7.136 9.136 1.8 Constraint E90.CA F56.CA 8.819 10.819 12.819 1.8 Constraint E90.CA F36.CA 13.516 15.516 17.516 1.8 Constraint P91.CA T3.CA 8.379 10.379 12.379 1.8 Constraint P91.CA F56.CA 12.344 14.344 16.344 1.8 Constraint P91.CA F36.CA 17.045 19.045 21.045 1.8 Constraint L92.CA T3.CA 11.224 13.224 15.224 1.8 Constraint L92.CA F56.CA 14.403 16.403 18.403 1.8 Constraint L92.CA F36.CA 18.748 20.748 22.748 1.8 Constraint S93.CA T3.CA 12.596 14.596 16.596 1.8 Constraint S93.CA F56.CA 14.826 16.826 18.826 1.8 Constraint S93.CA F36.CA 18.668 20.668 22.668 1.8 Constraint Q94.CA T3.CA 16.116 18.116 20.116 1.8 Constraint Q94.CA F56.CA 17.903 19.903 21.903 1.8 Constraint Q94.CA F36.CA 21.363 23.363 25.363 1.8 Constraint I95.CA T3.CA 17.453 19.453 21.453 1.8 Constraint I95.CA F56.CA 18.276 20.276 22.276 1.8 Constraint I95.CA F36.CA 21.059 23.059 25.059 1.8 Constraint I95.CA H8.CA 23.320 25.320 27.320 1.8 Constraint I95.CA E67.CA 32.694 34.694 36.694 1.8 Constraint Q94.CA H8.CA 22.322 24.322 26.322 1.8 Constraint Q94.CA E67.CA 31.922 33.922 35.922 1.8 Constraint S93.CA H8.CA 20.407 22.407 24.407 1.8 Constraint S93.CA E67.CA 28.612 30.612 32.612 1.8 Constraint L92.CA H8.CA 20.770 22.770 24.770 1.8 Constraint L92.CA E67.CA 27.428 29.428 31.428 1.8 Constraint P91.CA H8.CA 19.362 21.362 23.362 1.8 Constraint P91.CA E67.CA 24.368 26.368 28.368 1.8 Tue Aug 24 14:50:38 PDT 2004 Martina Koeva This is a brief summary on the attempts with the dimer (in the /dimer subdirectory). The try1 results looked like a terrible mess and I assume that was mostly the case, because the helix and strand constraints that str2/ehl2 predictions generated, clashed with what I had in mind for the structure. The predictions don't seem to find this sequence as a repeat and predict mostly helices for the second repeat/monomer segment (from 102 to 202). In order to force the structure that I want, I copied the t2k.dssp-ehl2 constraints from the try7.costfcn in the monomer directory and put them in try2.costfcn. I put in the same constraints twice, renumbering all residues to match the second segment of the new monomer. From what I can see in try2-opt1, the structure is slowly starting to get closer to the envisioned dimer. For try3 I will also attempt to put in distance constraints. Tue Aug 24 15:28:16 PDT 2004 Sol Katzman The result of dimerized try9s is not too bad. It has no clashes, and seems to line up residues K99-A101 with the N-terminal strand of the other monomer. (My intention was to get most of E90-I95 to that position, but T0263 does seem to have about 4 more residues on the C-terminus than 1lq9A). Tue Aug 24 15:53:04 PDT 2004 Kevin Karplus It seems that Martina never made a double-length model from one of the dimerized monomers to seed the dimer/decoys directory. I made one from Sol's dimer-1lq9-try9s model, and it score much better than any of the current models under dimer/ I'll start a dimer/ try3 run optimizing from existing models. Tue Aug 24 16:00:46 PDT 2004 Kevin Karplus oops. there is some mistake in from-dimer-1lq9-try9s. Let me try rebuilding it. Tue Aug 24 16:06:19 PDT 2004 Kevin Karplus OK, it is rebuilt, but I see that the dimer subdirectory is missing a lot of the t04 predictions (including the t04.many.frag library), so I'm rerunning make. Neither the t2k nor the t04 alpha predictions are particularly good in the second half, so I'm going to turn those weights down in the try3 costfcn. I'll also tweak the constraints a bit to give more weight to the new constraints that we are trying to form. Tue Aug 24 16:28:56 PDT 2004 Kevin Karplus I also picked up a lot of the d.58 folds from the t04 hits, and added them as MANUAL_TOP_HITS in dimer/Makefile, and am making extra_alignments, so that we have a richer set of fragments to work from. I note that many of the proteins have d.58 domains twice, so some may actually provide the dimerization internally. Tue Aug 24 17:13:29 PDT 2004 Martina Koeva > > It seems that Martina never made a double-length model from one of the > dimerized monomers to seed the dimer/decoys directory. > Actually, I did make a double-length model from the dimer-1lq9-try4 and I renamed it to dimer-as-monomer.pdb, but I kept it in the main directory. I did not start from it as a structure, but rather extracted its sheet and helix constraints, which I put in the /decoys directory. I put in the extracted sheet constraints for try1 and only added the helix constraints for try2. Hmm, maybe I should move the dimer-as-monomer.pdb file to dimer/decoys/, but the model is pretty similar to the later dimer-1lq9-try9s.pdb model. Tue Aug 24 17:49:04 PDT 2004 Martina Koeva Neat! Dimer try3-opt1 looks much more like it! It still curls up one of the ends into a helix (89-101), but the other end is slowly (190-202) slowly stretched out in the correct orientation (antiparallel) to the closest edge strand partner. Wed Aug 25 09:12:14 PDT 2004 Sol Katzman Still pursuing the scaffolding solution. Martina pointed out that in the dimer-1lq9-try9s model (as in her other previous dimers), the final C-terminal putative strand A98-A101 is parallel to the other monomer's sheet and we really want it to be antiparallel. Perhaps the extra residues that we have in T0263 (compared to 1lq9) form a hairpin at the C-terminus, and this 2-strand hairpin dimerizes to the other monomer's N-terminus. Then, what we want is to use the current locations of A101-A98 as the desired locations for P91-Q94. To make a hairpin, we can specify a sheet constraint, such as this: SheetConstraint P91 Q94 A101 A98 hbond 10 We can also use the current locations of Q97-I95 as the desired locations of A98-L100. (The current location of Q94 does not look right for A101). So I measured the current locations, and make the substitutions noted below. That will be try10s. In addition, I will do a try11s, which uses TryAllAlign rather than read-pdb.under, since we may be stuck in a local minimum. # Measured from try9s # Distance ALA101A.CA-THR3A.CA: 12.147 # Distance ALA101A.CA-HIS8A.CA: 22.391 # Distance ALA101A.CA-PHE56A.CA: 16.209 # Distance ALA101A.CA-PHE36A.CA: 18.967 # Distance ALA101A.CA-GLU67A.CA: 29.315 # # Distance LEU100A.CA-THR3A.CA: 10.470 # Distance LEU100A.CA-HIS8A.CA: 19.108 # Distance LEU100A.CA-PHE56A.CA: 14.136 # Distance LEU100A.CA-PHE36A.CA: 16.613 # Distance LEU100A.CA-GLU67A.CA: 27.337 # # Distance LYS99A.CA-THR3A.CA: 12.798 # Distance LYS99A.CA-HIS8A.CA: 19.655 # Distance LYS99A.CA-PHE56A.CA: 15.533 # Distance LYS99A.CA-PHE36A.CA: 16.907 # Distance LYS99A.CA-GLU67A.CA: 29.169 # # Distance ALA98A.CA-THR3A.CA: 12.544 # Distance ALA98A.CA-HIS8A.CA: 17.394 # Distance ALA98A.CA-PHE56A.CA: 14.278 # Distance ALA98A.CA-PHE36A.CA: 14.915 # Distance ALA98A.CA-GLU67A.CA: 27.573 # # Distance GLN97A.CA-THR3A.CA: 8.886 # Distance GLN97A.CA-HIS8A.CA: 16.450 # Distance GLN97A.CA-PHE56A.CA: 10.534 # Distance GLN97A.CA-PHE36A.CA: 11.651 # Distance GLN97A.CA-GLU67A.CA: 24.131 # # Distance ASN96A.CA-THR3A.CA: 9.793 # Distance ASN96A.CA-HIS8A.CA: 20.174 # Distance ASN96A.CA-PHE56A.CA: 11.548 # Distance ASN96A.CA-PHE36A.CA: 12.342 # Distance ASN96A.CA-GLU67A.CA: 25.940 # # Distance ILE95A.CA-THR3A.CA: 10.719 # Distance ILE95A.CA-HIS8A.CA: 22.758 # Distance ILE95A.CA-PHE56A.CA: 13.378 # Distance ILE95A.CA-PHE36A.CA: 14.861 # Distance ILE95A.CA-GLU67A.CA: 27.680 # 101,100,99,98,97,96,95(try9s locations) -> 91,92,93,94,98,99,100(new try locations) Constraint P91.CA T3.CA 10.147 12.147 14.147 1.8 Constraint P91.CA H8.CA 20.391 22.391 24.391 1.8 Constraint P91.CA F56.CA 14.209 16.209 18.209 1.8 Constraint P91.CA F36.CA 16.967 18.967 20.967 1.8 Constraint P91.CA E67.CA 27.315 29.315 31.315 1.8 Constraint L92.CA T3.CA 8.470 10.470 12.470 1.8 Constraint L92.CA H8.CA 17.108 19.108 21.108 1.8 Constraint L92.CA F56.CA 12.136 14.136 16.136 1.8 Constraint L92.CA F36.CA 14.613 16.613 18.613 1.8 Constraint L92.CA E67.CA 25.337 27.337 29.337 1.8 Constraint S93.CA T3.CA 10.798 12.798 14.798 1.8 Constraint S93.CA H8.CA 17.655 19.655 21.655 1.8 Constraint S93.CA F56.CA 13.533 15.533 17.533 1.8 Constraint S93.CA F36.CA 14.907 16.907 18.907 1.8 Constraint S93.CA E67.CA 27.169 29.169 31.169 1.8 Constraint Q94.CA T3.CA 10.544 12.544 14.544 1.8 Constraint Q94.CA H8.CA 15.394 17.394 19.394 1.8 Constraint Q94.CA F56.CA 12.278 14.278 16.278 1.8 Constraint Q94.CA F36.CA 12.915 14.915 16.915 1.8 Constraint Q94.CA E67.CA 25.573 27.573 29.573 1.8 Constraint A98.CA T3.CA 6.886 8.886 10.886 1.8 Constraint A98.CA H8.CA 14.450 16.450 18.450 1.8 Constraint A98.CA F56.CA 8.534 10.534 12.534 1.8 Constraint A98.CA F36.CA 9.651 11.651 13.651 1.8 Constraint A98.CA E67.CA 22.131 24.131 26.131 1.8 Constraint K99.CA T3.CA 7.793 9.793 11.793 1.8 Constraint K99.CA H8.CA 18.174 20.174 22.174 1.8 Constraint K99.CA F56.CA 9.548 11.548 13.548 1.8 Constraint K99.CA F36.CA 10.342 12.342 14.342 1.8 Constraint K99.CA E67.CA 23.940 25.940 27.940 1.8 Constraint L100.CA T3.CA 8.719 10.719 12.719 1.8 Constraint L100.CA H8.CA 20.758 22.758 24.758 1.8 Constraint L100.CA F56.CA 11.378 13.378 15.378 1.8 Constraint L100.CA F36.CA 12.861 14.861 16.861 1.8 Constraint L100.CA E67.CA 25.680 27.680 29.680 1.8 Wed Aug 25 11:06:11 PDT 2004 Sol Katzman It looks like we were in a local minimum for the monomer solutions because try10s (which read old models) did not do better than try9s, while try11s (which did TryAllAlign) did manage to put the desired residues somewhere near the sheet of the other monomer. We did not get the desired hairpin in try11s, however. It seems like my dimerization is a close second to (one half of) the full dimer model. Compare: T0263/dimer/T0263.try3-opt2.pdb.gz T0263/dimer-1lq9-try11s.pdb.gz Wed Aug 25 18:43:06 PDT 2004 Kevin Karplus I added a merged version of Sol's dimer-1lq9-try11s as dimer/decoys/from-dimer-1lq9-try11s.pdb.gz I also made monomers from two halves of dimer/try3-opt2, called decoys/from-dimer-try3A.pdb.gz and decoys/from-dimer-try3B.pdb.gz Superimposing these (from-dimer-try3-models.pdb) shows them almost identical except for the C-terminal end. Thu Aug 26 10:38:12 PDT 2004 Kevin Karplus The dimer/try4-opt2 is almost making the antiparallel strand for the C-terminal end. K200 and K39 are almost lined up. I split the double-length chain into two monomers (decoys/from-dimer-try4[AB]), superimposed them (from-dimer-try4-models.pdb), dimerized try4B (dimer-1lq9-try4B), and made the dimer into a double-length chain (dimer/decoys/from-dimer-1lq9-try4B.pdb) The new double-length chain meets the constraints better, but has too many clashes to score well with the dimer/try5.costfcn (even though that puts more weight on trying to form the new strand connection than try4 did). For the dimer/try5 run, I deliberately excluded the dimer/try4 models. so that from-dimer-1lq9-try4B would get polished. Thu Aug 26 13:44:35 PDT 2004 Kevin Karplus in dimer/try5, K39 is sticking out and interfering with the placement of K200, which is causing the whole strand to be misplaced. There are somewhat similar problems at the K99-K140 pair. For decoys/try6, I'll crank up the constraints and the weight on the new sheet constraints specifically, to try to really push the strands into place. The try6 costfunction still likes try5-opt2 best---it comes the closest to meeting the constraints. Thu Aug 26 14:08:01 PDT 2004 Martina Koeva I am going back to the alignment of the swapped strand to the sheet on the opposite monomer after a discussion with Kevin and Sol and I will attempt a different alignment: 37: LWKVW :41 (edge strand on the opposite monomer) | 201: LKAQN:197 (C-terminal strand on the original monomer) I am really not fond of the hydrogen bond between V40 and Q198, but it's worth a try. Thu Aug 26 16:22:34 PDT 2004 Sol Katzman With DeepView, we moved the C-terminal of each monomer closer to the sheet of the other monomer. Then, we used DeepView 'Strand' function to set the phi/psi angles of the residues N197-L201 for strand conformation, and positioned these next to the sheet of the other monomer until we got some Hbonding of the backbones. The final attempt that showed a good connection between N197-L201 and the other monomer was "dv5" from the following series of attempts: T0263.dv1.try5-opt2.pdb T0263.dv2.try5-opt2.pdb T0263.dv3.try5-opt2.pdb T0263.dv4.try5-opt2.pdb T0263.dv5.try5-opt2.pdb Thu Aug 26 17:15:10 PDT 2004 Kevin Karplus The hand-created models have a different alignment of the strand, so I put that SheetConstraint into try8.costfcn (which then scores dv3.try5-opt2 best). I'm doing an optimization from all existing models with the try8 costfcn. When it is done I'll split apart the dimer into monomers and be ready to submit (I hope). Thu Aug 26 18:58:10 PDT 2004 Kevin Karplus dimer/try8 finished, but it moved things around enough that redimerizing decoys/from-dimer-try8B does not line up the strands. That is--the previous edge strand of the sheet was detached somewhat in the optimization. I may need another round of optimization, with the other sheet constraints alomst as strong as the new ones. Thu Aug 26 19:54:59 PDT 2004 Kevin Karplus Assuming dimer/try9 works without tearing the sheet apart this time, I'll submit from-dimer-try9B (unless try9A looks better) try9-opt2 best unconstrained costfcn from-dimer-try4A.repack-nonPC best Rosetta repacked (as monomer) (unless try9A or try9B does better) try1-opt2 full-auto T0263-1lq9A-t04-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 best t04 template (2nd t02) ?? T0263-1iujA-t04-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 best t02 template (2nd t04) Thu Aug 26 22:50:01 PDT 2004 Kevin Karplus from-dimer-try9B is only slightly better than from-dimer-try8B when re-dimerized. We REALLY need to have a method that handles homodimers correctly! Fri Aug 27 09:03:00 PDT 2004 Sol Katzman Since dv5.try5B had slightly better connection of B-strand to A-sheet than try8B or try9B, I created decoys/from-dimer-dv5.try5B and then dimerized it. The result still does not hbond the B-strand to the A-sheet, but the positioning looks a little better than the try8B or try9B dimers. Thu Nov 18 23:45:12 PST 2004 Martina Koeva Based on the smooth gdt scores: best sam-t04 51.7466 (try1-opt1-scrwl) best submit 51.6053 model1 46.4347 auto 51.6050 align 43.3706 robetta best 53.0923 (robetta model5) robetta1 45.5741