Tue Jul  9 17:00:24 PDT 2002

9 July 2002 Kevin Karplus 

Strong hits to d.79.3.1 (1ck9A, 1jj2F, 1ffkE, 1e7kA, 1dt9A).
Probably homology modeling on 1ck9A, with full-length alignment.

Note: undertaker-try1 (without starting from an alignment) gets a
better score than undertaker-try2 (which does start from alignments).

Perhaps we should seed an undertaker script with all the 1ck9A
alignments, to give more variety than just the 2track-protein-STR
alignments?


11 July 2002 Kevin Karplus

Looking at T0150.try1-opt.pdb and T0150.try2-opt.pdb with rasmol, it
looks like there may be some misalignment---the predicted strands are
off by one or two residues.  The gapless alignment to 1ck9A looks
pretty convincing at conserved residues though, so I'm a little
reluctant to do any editing.

I have put together an undertaker script that tries all the alignments
of 1ck9A, 1jj2F, 1ffkE, 1e7kA, and 1dt9A, in the hope of getting a
slightly better packing.

T0150.try3-opt.pdb scores a little better, and looks quite compact,
but I may want to add some constraints for the final strand to make it
Hbond properly.  Strand 1 may be peeling up a tiny bit also, though
that may be from the original template.


12 August 2002 Kevin Karplus

Reran make with new template library---no changes to top hits.
Modernized define-score.script and undertaker.script and reran without
adding constraints.

17 Aug 2002 Kevin Karplus

Best score currently is try4-opt-scwrl.
It's not bad, though a bit foamy.

I upped the weight for dry6.5, dry8, and break, and try4-opt-scwrl is
still best.

18 Aug 2002 Kevin Karplus 19:58

Best score now try5-opt.  It does not look bad, other than what seems
to be an off-by-1 (or 2) on the predicted positions of the beta
strands.

It looks like I could get another strand on the sheet by adding Hbonds
V17.N-A84.O and V17.O-A84.N.  Adding these constraints (as
try6.constraints), still leaves try5-opt scoring best.

Let's try re-optimizing try5-opt with these constraints added to the
score definition.

Mon Aug 19 13:19:48 PDT 2002	Kevin Karplus

New best is try6-opt, a slight improvement over try5-opt, but not
getting the extra strand.

I'm a bit worried about the possiblity of misalignment, so I've added
the "str-heavy" alignments and am going to try running from the
beginning again (no seed conformations) for try7.


Mon Aug 19 16:39:13 PDT 2002	Kevin Karplus

try6-opt still best scoring, but try7-opt also looks good to me, and
it DOES get the extra strand.

Let's try polishing try7-opt

21:27  Kevin Karplus
try8-opt beats try7-opt, but still loses to try6-opt.
I tried tweaking the scores to make try8-opt look better, but there is
one bad break (hear the end of the chain) that messes things up.
I'll try doing a long polishing run tonight with lots of
crossandinsert, in the hopes that the good parts of try6-opt and
try8-opt can be combined.

Tue Aug 20 13:27:35 PDT 2002	Kevin Karplus

Best score now T0150.try9.10.100.pdb, but it is much foamier than I'd like.

If we turn down clash and break penalties somewhat, and turn up dry6
and try8, the best remains unchanged.  We might as well submit it, as
the improvement over try6 is small.

I'll submit try9.10.100 as first model and try8.7.100 as second model.