Fri Aug 16 16:04:46 PDT 2002 t0192 Homology model 1i1dA d108.1.1 Aug 26, 2002, Monday Oscar Hur There are some breaks in try1. Reduced Break is set to 15 and run try2. Aug 27, 2002, Tues Oscar Hur Change the parameters to see if the gap is reduced CloseGap to 10 JiggleSubtree to 0.5 OptSubtree to 0.5 Run try3 Aug 28, 2002 Wed Oscar Hur Add the value of OneRotamer 0.2 InsertFragment change to 10. Run try4 Aug 29, 2002, Thur Oscar Hur Big breaks still exist in the pairs of amino acid residues: 15-16, 21-23, 28-29, 137-138, 154-155 Change of following values to: CloseGap 25 OptSubtree 2 JiggleSubtree 2 InsertFragment 20 InsertSpecificFragment 20 And Run try5 Aug 30, 2002 Fri Oscar Hur Big breaks in 15-16, 18-19, 27-28, 41-42, 83-84, 120-121 Increase InsertFragment to 50 InsertSpecificFragment to 50 2 Sept 02 10:13 Kevin Karplus Modified define-score.script to increase burial, break, and clashes parameters to ones that have worked ok on other homology models. Best score is now try6.13.60 This has a number of awkward mid-helix gaps, and the C-terminal strands S124-V129 and S146-E152 have gotten detached from the sheet. We'll need to add constraints and try again from the initial alignments. Let's guess S124-V129 parallel to L89-M97 and to S146-E152. 84> wigkLLYLEDFFVMsdy 120> crcsSMHFLVaew 143> rgaSDLSSEegw In the current model, C54 comes close to C120 and C122. If there is a disulfide here, or a metal-binding site, it would be difficult for S124-V129 to move into the sheet. UNLESS the C-terminal strands are on the other side of the sheet, with S124-V129 antiparallel to K3-P8. 10< tapRIVFKam 120> crcsSMHFLVaew 143> rgaSDLSSEegw Let's try this set of constraints, with maybe_metal turned on to encourage clustering of the 3 CYS. With these constraints (try7.constraints), the best score is for try6.15.60 Let's start over from scratch with these constraints, putting the pseudocounts for the various methods back to more normal settings, and making this be a short run, so I can look at the results and decide whther they are worth doing more polishing on, or whether we have to choose a different set of constraints (like putting the C-terminal strands on the other side of the sheet). Let's try the other arrangement 84> wigkLLYLEDFFVMsdy 120> crcsSMHFLVaew 143> rgaSDLSSEegw as try8.constraints, and do a run starting from those constraints also. We can see which of the two is more convincing. Note: try6.15.60 scores best with the try8.constraints also. 13:17 try7 and try8 have finished their short runs. TRY7: C-terminal strands at beginning of sheet With the try7 constraints, try7-opt is the new best score. It does not look good as S150-E152 has wound up into a helix and S124-V129 is at right angles to the sheet. I has paired up C54 and C14 instead of C54-C120-C122, For try9, let's turn off maybe_metal, and make C54-C14 an Sbond, and turn on mave_ssbond for C12-C122 Let's also adjust alignment to 10< tapRIVFKam 120> crcsSMHFLVaew 143> rgaSDLSSEegw and add a straightness constraint for S146-E152 With these constraints, the best-scoring decoy is still try7-opt. Let's reoptimize from there. TRY8 C-terminal strands at end of sheet: With the try8.constraints, the best score is still try6.15.60, but try8-opt looks like it could make it, with a little fixing of the constraints. If the strand K87-D99 is not an edge strand, then the ED pair in the middle will probably cause a bulge, as try8-opt is doing. The problem is that the orientation at the ends can't really be flipped. The CRC at C120-C122 has to have the Cys facing each other and the Arg pointing out, which puts the C120 on the side of the sheet that the helix packs against. This makes the exposed side be (uppercase) CsSmHfLvAe, which should be ok, since the E is at the end of hte strand and can stick out the other side of the sheet. I'm having trouble coming up with a set of constraints that would make a sensible sheet here---let's hope that the try9 constraints work! 2 Sept 2002 17:03 Kevin Karplus In order to get the C-terminal strands close and make the disulphides, the first strand was pulled out of the sheet. The final strand is still not straight! If we staple that strand down aligned as follows: // 51> fyHCLVAEv // 10< tapRIVFKam // 120> crcsSMHFLVaew // 143> rgaSDLSSEegw and remove the ssbond scores, we still get try9-opt as the best. Let's try reoptimizing from there Wed Sep 4 13:01:59 PDT 2002 Kevin Karplus Oscar said in group meeting today that he thought t0192 was almost done, though there are no notes on it here in the README file. The current best score (with try10.constraints) is try10-opt, but it is a terrible mess, with only 2.5 strands in the sheet and S146-E152 wound into a helix. The C14-C54 disulphide has some problems---the sulfurs aren't coming close. We may need to break at A12 to get better placement of teh N-terminal strand. Let's try shifting the alignment to // 51> fyHCLVAEv // 10< tapRIVFKam // 120> crcsSMHFLVaewne // 143> rgaSDLSSEegw With these (try11) constraints, try10-opt is still best. Let's try reoptimizing from there. Wed Sep 4 18:02:58 PDT 2002 In try11-opt strand S124-V129 is at right angles to strand F4-P8. There is a bad kink at y78, making a bad gap between L89 and Y90. This needs more thought and more optimization--- how should the sheet be built? Sep 5, Thur 3:58 PM Oscar Hur There are big breaks in 71-72 and 89-90 Add constraints to the two strands and change the orientation a little to see if the gaps improve. Also at the same time, try to pack strands together. // 4> FVIR // 58< EAVLCH // 70> sIVGFAMYYFt // 96< VFFDELYLL Do not know why 148-151 and 154-155 were packed in helix. Run try12. Sep 6th, Oscar Hur, 9:47 AM I do not see my try12 run. I will try to do try12 again. Sep 6th, Oscar Hur, 2:02 PM Look at try12-opt-scwrl The constraints apparently have not worked to pair the beta strands. I will increase the weight of the constraint to 0.5 in define-score. I will also change the constraints from c-beta to hydrogen bonding. Run try13 7 Sept 20002 Kevin Karplus I'm not too surprised that turning the constraints way up (try13-opt) collapsed the structure into a terrible mess. The secret to using constraints is to guess the right ones, not force the wrong ones. Even with constraint weight turned back down to 0.1, the best-scoring model is try13-opt, but it looks terrible. Going back to the try11.constraints, the best-scoring is still try11-opt. With no constraints, the best scoring is try6.13.60 Let's try giving that one to VAST, and seeing whether there is a larger protein that it might be part of. Job VS31332 password casp5t0192 We have large matches to 1qsmD and 1i1dD---82 and 101 residues, both extendible to almost complete matches. We can change these to 1qsmA and 1i1dA---it doesn't look like this protein is SUPPOSED to form a single sheet. Those two C-terminal strands have that unusual arrangement in the crustal structures. So we may really be done after all. The Hbonds in the two vast alignments were extracted and made try14.constraints. 21:32 Kevin Karplus Hmm, optimizing with the try14 constraints moved the final, unattached strand so that it passes though the gap where the parallel strands separate, rather than around the outside as in the templates. At the other end of the protein, C54 andF46 clash badly. 23:22 Maybe I should add some more constraints to hold the last strand roughly in place---maybe contacts between R142.CG and F127.CE2 and Y140.CZ and F127.CZ or Y140.CE1 and V129.CG1. I just noticed that I hadn't started try14 from a completely random initial starting conformation---let's fix that in try15. 23:45 The extra constraints make try14 score poorly (as hoped), and the try15 run seems to be reducing the score nicely---let's see if it ends up ok. 8 Sept 02 00:11 Kevin Karplus Well, try15 didn't make the SAME mistake as try14, but try15-opt has curled 146-152 up into a helix. maybe I should try to place the Cterminal helix in the "right" place to keep the strand stretched out, and I should add length constraints. 8 Sept 02 09:05 Kevin Karplus try16-opt isn't what I expected, but I like it. Let's take out the (probably incorrect) constraints on the position of the final helix and reoptimize. 12:08 Kevin Karplus Hey--cool, we're beginning to get some extra Hbonds to attach the final piece to the split sheet. Let's try to encourage that.