Wed Jun 18 09:15:57 PDT 2008
T0460
Make started Wed Jun 18 09:17:14 PDT 2008
Running on peep.cse.ucsc.edu

Wed Jun 18 20:11:57 PDT 2008 Kevin Karplus

ORFan, new-fold or very remote homology. (best hit E-value 31)

The try1-opt3 model is very gappy---I need to increase breaks to close
things up.


Wed Jun 18 20:22:42 PDT 2008 Kevin Karplus

try2 started from alignments, but with costfcn that penalizes gaps and
clashes more.

Wed Jun 25 09:02:06 PDT 2008 Kevin Karplus

MQAU quality assessment favores SAM-T08-server, Zhang-Server, and BAKER-ROBETTA
MQAC quality assessment favores SAM-T08-server, Zhang-Server, METATASSER, and pro-sp3-TASSER

metaserver runs with try1 costfcn have been started.

Sat Jun 28 14:40:53 PDT 2008   SAM-T08-MQAO hand QA T0460 Submitted
Sat Jun 28 14:40:53 PDT 2008   SAM-T08-MQAU hand QA T0460 Submitted
Sat Jun 28 14:40:53 PDT 2008   SAM-T08-MQAC hand QA T0460 Submitted

Sat Jul 12 12:17:22 PDT 2008 Kevin Karplus

MQAC1 and MQAU1 runs both chose SAM-T08-server_TS1.
Perhaps I should do MQAX1 and MQAX2 runs that exclude the SAM servers.

Sat Jul 12 12:27:49 PDT 2008 Kevin Karplus

MQAX1 and MQAX2 started.

Actually, there are some things I like about the MQAC1 and MQAU1
predictions (from SAM-T08-server_TS1).  There are several sheets that
don't quite form, but would be quite nice if they did.  Perhaps adding
some sheet constraints would fix things up.

Sat Jul 12 13:03:57 PDT 2008 Kevin Karplus

I tried making up some sheet constraints for try3.costfcn, but I did
not come up with a consistent set.

I'll try using it to optimize the MQAC1 and MQAU1 models anyways.

Sat Jul 12 15:26:25 PDT 2008 Kevin Karplus

Interestingly, the best scorers with the try3 costfcn was not
try3-opt3, but MQAX2-opt3 and MQAX1-opt3, both from optimizing BAKER-ROBETTA_TS5.

Those models look pretty good, but quite different from the other
models I've been looking at.

Perhaps I should do a run from alignments with the MQAX2 constraints.

Sat Jul 12 18:17:25 PDT 2008 Kevin Karplus

MQAX2 and MQAX1 score best with try4.costfcn, and try4 doesn't even do
as well as try2 and try3.

try4 has some difficult gaps to fill, but the sheet looks fairly plausible.

I'll try polishing it in try5.

For try6, I'll try polishing try2/try4 using the same costfcn as for try5.

Sat Jul 12 20:10:25 PDT 2008 Kevin Karplus

try5 gets a bit better than try6, though both have bad n_ca_c values,
which gromacs makes even worse.  They have bad breaks also.

Sat Jul 12 20:22:00 PDT 2008 Kevin Karplus

I currently have 4 different lineages of models that don't have much
in common:

	MQAX2-opt3	from BAKER-ROBETTA_TS5
	try3-opt3	from SAM-T08-server_TS1 (and try1)
	try5-opt3	< try4-opt3 < 2ictA?
	try2-opt3	< 2j1dG?

Rosetta, naturally, likes MQAX2-opt3.gromacs0.repacck-nonPC best.

I'm not convinced by any of them, so I should probably submit all 4.
Question: do I do more gap closing on the ones with big breaks?

Wed Jul 16 13:26:20 PDT 2008 Kevin Karplus

I'll do one more metaserver run, with high beta_pair score and high
neural-net property scores, excluding SAM servers (and
BAKER-ROBETTA_TS5), and see what is picked up.

Since I have no clue what this protein should look like, I'll submit 5
junky models.

Wed Jul 16 17:30:08 PDT 2008 Kevin Karplus

I don't see anything reasonable in any of my models or ther metaserver models.
The two optimized from the BAKER-ROBETTA models are as good as I'm
likely to get.  I'll submit

ReadConformPDB T0460.MQAX7-opt3.gromacs0.repack-nonPC.pdb	# < BAKER-ROBETTA_TS4
#best rosetta energy

ReadConformPDB T0460.MQAX2-opt3.pdb	# < BAKER-ROBETTA_TS5

ReadConformPDB T0460.try3-opt3.pdb	# < MQAC1-opt3 < SAM-T08-server_TS1

ReadConformPDB T0460.try5-opt3.pdb		# < try4-opt3 < align(2j1dG)

ReadConformPDB T0460.try2-opt3.pdb		# < align(2j1dG)

The MQAC quality assessment thought that none of the servers got up to
30% GDT, and I'm inclided to agree.