Thu Jun  8 08:36:20 PDT 2006
T0323
Make started Thu Jun  8 08:36:51 PDT 2006
Running on camano.cse.ucsc.edu

Mon Jun 26 22:27:56 PDT 2006 Kevin Karplus

This one is due this week, and no one has looked at it yet!

Luckily, aside from a few residues at the termini, the automatic
try1-opt2 model looks pretty good.

I'll do a polishing run, then probably submit.

try2 started on shaw.

Tue Jun 27 10:06:53 PDT 2006 Kevin Karplus

try2-opt2 has a substantial improvement in cost, though it still has a
few very minor breaks.  Rosetta likes it better than try1-opt2 for
repacking, but is not really fond of either.


I wonder it it is worth polishing further, or if I should create a new
model from the initial alignments, trying to get different termini?

Tue Jun 27 10:23:16 PDT 2006 Kevin Karplus

try3 uses the try2 costfcn, but starts over from alignments.
It also does two different random starts, to try to get a bit more
variety in the termini.

try3 started on camano.

Tue Jun 27 11:56:54 PDT 2006 Kevin Karplus

try3 does not get even as good as try1.

Tue Jun 27 12:23:02 PDT 2006 Kevin Karplus

I guess it is time to polish.

try4 (polishing run) started on camano.

Tue Jun 27 14:56:46 PDT 2006 Kevin Karplus

try4 done.  try4-opt2 is best with try4 costfcn, unconstrained
costfcn, and is the model that rosetta best likes repacking.

There are still one or two moderate breaks:
	T0323.try4-opt2.pdb.gz breaks before (T0323)F4 with cost 0.659212
	T0323.try4-opt2.pdb.gz breaks before (T0323)V91 with cost 0.446052

There is probalby not much point in further polishing, though, so I
think I'll submit

ReadConformPDB T0323.try4-opt2.pdb
ReadConformPDB T0323.try1-opt2.pdb
ReadConformPDB T0323.try3-opt2.pdb
ReadConformPDB T0323.undertaker-align.pdb model 1	# from 1ornA
ReadConformPDB T0323.undertaker-align.pdb model 2	# from 1mun

Tue Jun 27 15:07:53 PDT 2006 Kevin Karplus

So submitted.

Mon Aug 21 16:31:54 PDT 2006 Kevin Karplus

Our best model is try3-opt2.gromacs0, but align1 has a much higher GDT.
Closing the big gaps seems to have moved things around too much---this
might be an argument either for frozen core or for picking up
helix-packing constraints from alignments.


Fri Sep  1 14:35:34 PDT 2006 Kevin Karplus

We probably worked from the wrong alignment here.  The align3 model
was our best with a GDT of 50.5%, and the best we submitted was model4
(GDT 40.8%) which was  align1.  We should have submitted one fewer
polished models and one more alignment.

Sun Sep 10 08:41:49 PDT 2006 Kevin Karplus

Our best model was model4 (scwrl on align1), but SAM_T06_server_TS4
(scwrl replacement on align3) is much better.

MetaTasser had the best model of any of the servers.

Fri Mar 23 15:45:33 PDT 2007 Kevin Karplus

align3 was our best alignment on both domains.

For domain 1, try4-opt1-scwrl was was slightly better than the best we
submitted (model1=try4-opt2), as was try2-opt2.gromacs0.

For domain 2, our best model was model1=try4-opt2 (well,
try-opt2.repack-nonPC was slightly better).