Tue Jun 25 14:25:25 PDT 2002 T0142 Clear homology model to 1i9zA, 1i9yA, 1ako, 1bix, ... (d.151.1.2), with 1i9zA/1i9yA as obvious template. Note: 1i9yA and 1i9zA are ony 0.2 Angstroms RMSD apart over 336 residues. FSSP chose 1i9yA as representative, but Dunbrack chose 1i9zA. Let's remake with tighter BEST_EVALUE (say 0.02). 26 June 2002 Kevin Karplus try1-opt seems to be based on a model that had a 1i9zA insertion most recently, with no further improvements in the super-iteration. It looks moderately good, though the strand L95-D110 has been pulled off the sheet. Perhaps we should try a structural alignment to 1i9zA, to get the alignment to use as a starting point. 29 July 2002 Jonathan Casper Adding SCWRLing and pred_alpha to the undertaker script. 10 Aug 2002 Kevin Karplus With the rather out-of-date score function currently defined, the best score is try3-opt. After adjusting the parameters and replacing radius_fit by phobic_fit, the best is try2-opt-scwrl Let's start over with more alignments (make try-alignments) and the new scoring function. Let's also run the optimization for longer, to see what comes up. If it looks good, we might want to do a VAST alignment with 1i9yA or 1i9zA to get a better alignment to start from. 11 Aug 2002 Kevin Karplus Reran 'make' with new template library. No changes among top hits. 12 Aug 2002 Kevin Karplus try4 run is still going---it takes a long time to pack something this big. When it finishes, I'll have to look at the results and try one more run (somewhat shallower), perhaps upping CloseGap. 13 Aug 2002 Kevin Karplus try4-opt is new best score. The sandwich looks pretty good, except for the detatchment of K253-L258, and some small breaks. Let's add a constraint Hbonding I255-T281 P257-R279. Oops---can't Hbond the N of the proline! So let's just do the 255-281 bonds, and do CB proximity for L258 and F278. 15 Aug 2002 Kevin Karplus The try5 undertaker run took too long--probably need to reduce the number of generations if I do another run. try5-opt-scwrl is the new best score, but I haven't looked at it yet. 20:55 Kevin Karplus The try5-opt-scrl structure looks pretty good. There are a few breaks that need attention, so let's do the next run with the break penalty turned up. Fri Aug 16 11:42:30 PDT 2002 try6 is still running. Current best is try6.4.60. It has about 4 more hours to run. Fri Aug 16 16:51:50 PDT 2002 Hmm---it took longer than I thought. try6-opt-scwrl is new best score. There is a bad break at L166-F167, also S121-F122, both in strands. There is another break at A237-T238, but that's a loop region. There were improvements in several breaks relative to try5-opt-scwrl, but several others are unchanged. The packing is not as tight as I'd like, but I'm not sure we'll get enough improvement to justify another long run, but I'll do another anyway, with the break penalty turned up even higher 17 Aug 2002 07:15 Kevin Karplus try7 is still running. try7.9.80.pdb is new best. 08:10 try7 took 14 hours to complete, and try7-opt is new best. There were tiny changes in the breaks, but nothing to get excited about. I think it is time to submit.