Wed Jul 17 12:49:24 PDT 2002 t0158 17 July 2002 Kevin Karplus Obvious homology to 1jjiA, strong similarity to many similar templates. 18 July 2002 Kevin Karplus There seems to be a disulphide bond between C254 and C282, and there could be one for C78 and C116. Less likely, but still possible is one between C142 or C143 and C185. Note: 1jjiA does NOT have disulphides. I accidentally killed the try2 run before it completed. The best score from that run was for T0158.try2-al8.7.20.pdb, which has pulled one strand out of the middle of the sheet. I think the disulphide bonuses are much too big. Reducing the weight of the disulphide cost pulled try1-opt-scwrl to the top again. I'll try re-running with the new score function. 19 July 2002 Kevin Karplus try3-opt-scwrl is the new best. It looks pretty good except for * the bad break between D73 and T74, * the small break between G211 and V212, * the small break between A85 and T86 * the small overlap between T248 and R249 * the predicted strand V72-C78 is not H-bonded (and is broken) * the predicted strand T60-P66 is wound up into a helix The problems for V72-C78 may be caused by the high score for the disulphide V78-C116. I could try upping the penalty for breaks and lowering the weight of the disulphide bonus. try4-opt-scwrl scores best on the new scoring function. There are still the same 4 nasty breaks (particularly the one between D73 and T74), though they have been reduced somewhat from try3-opt-scwrl. Perhaps I should add some CB constraints between the two strands that are getting messed up to force a hairpin. The trouble is, I can't figure out a decent pairing---maybe 59> ATRAYMVP tky |||| 79< pCFLRTE vqg Looking at the T0158.t2k.undertaker-align.pdb file, particularly at model 12 (T0158.t2k-2track-undertaker.a2m:1jjiA), I see a perfectly fine hairpin 59> ATRAYMVPTk 79< pCFLRTEVQGy with Hbonds on 419 N A59 O C78 579 422 O A59 N C78 575 431 N R61 O L76 562 440 O R61 N L76 556 447 N Y63 O T74 543 457 O Y63 N T74 538 467 N V65 O V72 527 472 O V65 N V72 522 481 N T67 O G70 511 486 O T67 N G70 509 One problem with this alignment is that C78 ends up on the opposite side of the sheet from C116, which would make the disulphide there impossible. So--which is more likely, that there isn't a diulphide there or that the alignment is wrong? All the alignments to 1jjiA agree on almost all the alignment, disagreeing slightly on one or two residues where there are gaps (such as D246 L247). Since T0158 has an insertion near C78 (A85T86 replacing V), but I see no way to make either C78 or C116 flip to the other side---the beta strands have a lot of identity and moving off-register by 1 would change the hyrdophobicity pattern. Let's turn off the C78-C116 disulphide and try again. 23 July 2002 Kevin Karplus After try5, the best-scoring decoy is try5-opt.pdb. It looks like the strand K2-P8 maybe should be next to H278-Y287, but I'm not certain enough of this to put it in as a constraint. After replacing the alpha and alpha_prev cost functions with the new pre_alpha2 cost function, the best-scoring decoy is still try5-opt. Let's try adding another possible SSBOND (C143-C185) to see if it stabilizes the loop around C185. With this added to the scoring function, try5-opt still comes out on top. 23 July 2002 Kevin Karplus T0158.try6-opt is the new best scorer, but it is only slightly better than try5-opt. The C143-C185 bridge is NOT formed---in fact these residues are identically spaced as in try5. Aside from deciding what to do with the C-terminus, I'm not sure what else we should do with this file. 18 Aug 2002 Kevin Karplus Let's modernize the score function and the undertaker script and start over from the beginning. Mon Aug 19 13:45:41 PDT 2002 kevin Karplus T0158.try7-opt is new best score. There are pretty bad breaks mid-sheet after E283 and F255, probably because of the pair of CYS at 254-282. Rasmol does not see either of the disulphide bridges as being formed (wrong rotamers, perhaps). If we rescore with no cost function for ssbonds, try7-opt still scores best. Let's try to close up the gaps on this version, uppping the break penalty, and making the packing parameters (dry5-dry8) more heavily weighted. 19 Aug 19:35 Kevin Karplus try8-opt looks pretty good. Let's submit it. I'm giving up on the gaps, just so I can this out the door.