Wed May 14 09:07:37 PDT 2008
T0401
Make started Wed May 14 09:08:01 PDT 2008
Running on cheep.cse.ucsc.edu

Wed May 14 10:36:07 PDT 2008 Kevin Karplus

No good hits with blast, but the HMMs have pretty confident hits for
three templates not in SCOP:

2prvA	153	1.13e-15
2icgA	160	4.75e-15
2pagA	135	1.67e-13

These are identified as "SMI1 / KNR4 family. Proteins in this family
are involved in the regulation of 1,3-beta-glucan synthase activity
and cell-wall formation" in the NCBI Conserved domain database.

Wed May 14 12:53:16 PDT 2008 Kevin Karplus

The try1-opt3 model and the three models from alignments are in fairly
close agreement.

I think we want to extend the helix starting at F118 out to Q134 and
maybe to V136, since one of the templates (2prvA) has a helix out to Q134.
The predictions for a helix N-capping motif at NFN (starting at N129)
looks pretty strong though, I'm torn about whether to extend the helix
past L128.


The strand starting at G44 may need to be extended a residue or two

SheetConstraint (T0401)G44 (T0401)G46	(T0401)I55 (T0401)I53	hbond (T0401)G44	1
SheetConstraint (T0401)W8 (T0401)D12	(T0401)G46 (T0401)N42	hbond (T0401)Y9	1

changed to 

SheetConstraint 	G44 	K47		I55 	Y52	hbond 	G44	1
SheetConstraint 	K7 	D12		K47	N42	hbond 	Y9	1


Wed May 14 16:53:07 PDT 2008 Kevin Karplus

Both try1-opt3 and try2-opt3 have pretty bad breaks.
try1-opt3 meets align_constraints better, try2-opt3 meets the
try1.constraints better.

I think that I need to balance the costs between the two a bit better,
then do a polishing optimization to close the gaps---perhaps an
optimization that relaxes the clash penalty to close gaps, then
tightens the clash penalty.


Before I  try a polishing run, I'll do another run with just the 3
good templates, and with the cost function that favors closing gaps.

Wed May 14 18:48:55 PDT 2008 Kevin Karplus

try3-opt3 has much smaller breaks but somewhat worse clashes than try2-opt3.
gromacs improves somewhat on the total cost, by reducing clashes at
the expense of increasing breaks.  Rosetta likes the
try3-opt3.gromacs0.repack-nonPC models best so far.


Wed May 14 20:50:13 PDT 2008 Kevin Karplus

try4-opt3 reduces clashes substantially, but increases breaks a bit.

Mon May 19 22:37:35 PDT 2008   SAM-T08-MQAO hand QA T0401 Submitted
Mon May 19 22:37:35 PDT 2008   SAM-T08-MQAU hand QA T0401 Submitted
Mon May 19 22:37:35 PDT 2008   SAM-T08-MQAC hand QA T0401 Submitted

Fri May 23 09:05:43 PDT 2008 Kevin Karplus

MQAU and MQAC like the SAM-T08-server and the Zhang-Server

I started the metaserve scripts with try3 as the costfcn on the moai cluster.

Thu May 29 14:26:48 PDT 2008 Kevin Karplus

MQAU1 and MQAC1 are both based on SAM-T08-server_TS1.  I should
compare to Zhang-Server_TS1 (best according to MQAC) to see if I'm missing anything.


Thu May 29 14:35:39 PDT 2008 Kevin Karplus

There are some differences between try4-opt3 and the Zhang-Server model.
The biggest one is probably residues N129-V136, which Zhang continues
as a helix (as do the MQAU-opt3 and MQAC-opt3 models) and which
try4-opt3 bends around to pack a C-terminal helix. I think I like
try4-opt3 beter (for the 2ry prediction and the burial).  I should
look at the templates and see if they all have a long helix here.

But for I108-K112, I think I like the Zhang-Server model better.

The Zhang-Server model may also pack the conserved F81 and Y34
better---again, I need to look at templates.

Thu May 29 17:59:02 PDT 2008 Kevin Karplus

There is no evidence I see in the templates of any special packing of
the aromatics---whatever provides the tightest fit should be chosen.

2prvA has a longish C-terminal helix sticking out, contributing to a
dimeric interface. (2pagA appears to be monomeric, though 2icgA is
claimed to be hexameric by PQS at http://www.ebi.ac.uk/msd-srv/pqs/).
The C-terminal helix is anti-parallel to an N-terminal helix.  This is
what has been copied for the Zhang-Server model, but I think I like
try4-opt3 better. 

I'll make a chimera that takes I108-K112 from the Zhang-Server_TS1
model into try4-opt3, and optimize this to make try5.
After that, I should probably do a polishing run.

Thu May 29 19:09:03 PDT 2008 Kevin Karplus

try5-opt3 doesn't score quite as well as try4-opt3, but they are
pretty close (and rosetta likes try5-opt3.gromacs.repack-nonPC best).

Thu May 29 19:17:47 PDT 2008 Kevin Karplus

I think I'll give up on this one---I've reached the point of
diminishing returns.

I'll submit

ReadConformPDB T0401.try5-opt3.pdb
ReadConformPDB T0401.try4-opt3.pdb
ReadConformPDB T0401.MQAC1-opt3.pdb
ReadConformPDB T0401.try1-opt3.gromacs0.pdb
InFilePrefix 
ReadConformPDB T0401.undertaker-align.pdb model 1	# from 2prvA