Mon Jul 15 20:30:17 PDT 2002
t0154

Obvious homolog of 1ihoA (pantoate--beta-alanine ligase).

16 July 2002 Kevin Karplus

The models from alignments look good (particularly model 5 for
T0154.t2k-2track-undertaker.a2m:1ihoA), but try1-opt is terrible,
because the define-score.script file had a number of inappropriate
contraints inherited from a different target.  Trying again without
the bogus constraints.

The try2-opt structure looks pretty good, though there are still a few
small breaks.  I'll try running it through SCWRL, reoptimizing, and
running through SCWRL again.

The try3-opt structure looks good also---breaks have been slightly
reduced but there are more clashes than in try2-opt-scwrl.  Increasing
the penalty for breaks a bit, I'll try running again with the
try2-opt, try2-opt-scwrl, try3-opt, and try3-opt-scwrl conformations
as starting points.

17 July 2002 Kevin Karplus

The try4-opt-scwrl structure looks good, but has not improved much
over try3-opt (mostly sidechain and clash scores).  I wonder if it is
worth doing another round.

Fri Aug  2 15:08:19 PDT 2002 Kevin Karplus

Updating the score function to include pred_alpha2 and phobic_fit
makes earlier iterations of the try4 run score best.

Let's reoptimize starting from try3-opt and try4-opt and the top 3
decoys with the new score function.

3 Aug 2002 Kevin Karplus

try5-opt-scwrl is the new best scorer.  It looks pretty good.
There are several small breaks, so it may be worth re-optimizing from
here, turning up the break and clash penalties.

4 Aug 2002 Kevin Karplus

try6-opt is the new best score, with try6-opt-scwrl close behind.
I think the breaks are now too small to worry about, and any more
fussing is as likely to make things worse as better.  I'll submit
try6-opt.