Wed Jul 12 13:58:57 PDT 2006 T0372 Make started Wed Jul 12 14:03:25
PDT 2006 Running on halo.cse.ucsc.edu

Thu Jul 13 10:38:45 PDT 2006 Kevin Karplus

T0372 seems to a 2-domain protein, with the domain boundary somewhere
around E116-D122.  The C-terminal domain (d.108.1.*) is pretty well
captured by the whole-chain modeling, but the N-terminal domain may
need work.

I started M1-D122 on the farm cluster.

Thu Jul 13 10:47:38 PDT 2006 Kevin Karplus

Several of the sequences in the t06 multiple alignment are labeled as
membrane proteins.  We'd better run through TMHMM---membrane proteins
are a whole different ball game.

http://www.cbs.dtu.dk/services/TMHMM/ sees no TM helices.
http://www.cbs.dtu.dk/services/LipoP/ doesn't see lipoportein or
signal peptide.

OK, I'm going to assume that this is *not* transmembrane or even
membrane-anchored, though it may bind to a membrane protein.

Thu Jul 13 11:11:49 PDT 2006 Kevin Karplus

There are enough sequences in the multiple alignments to have some
confidence in the secondary and rr predictions.


Thu Jul 13 12:01:14 PDT 2006 Kevin Karplus

I just noticed that T0372 and T0374 seem to share the d.108.1.*
domain.  How similar are they?

Make started Wed Jul 19 09:05:57 PDT 2006 Running on
orcas.cse.ucsc.edu


Wed Jul 19 09:06:47 PDT 2006 Chris Wong

Prof. Karplus fixed a bug in make.main that reduced the number of
alignments produced.  I just started (make -k >& make2.log; gzip -9f
make2.log) & to get the additional alignments for this target.
Running on orcas.


Wed Jul 19 13:21:02 PDT 2006 Chris Wong

Looked at some alignments, trying to figure out what to do with this
one.  First, there is an intrachain disulfide between C176 and C181
that is described in the doc.html for this target.  We should specify
this in the costfcn.

I made a try2.costfcn, and uncommented <known_ssbond 0.5> and added
<ssbond C176 C181>.

Next,"crystal structure indicates dimeric oligomerization state."  I
see that Prof. Karplus has made a subdomain 1-122 and ran a make on
it.  It has in the top hits 2D4P, which looks a lot like one half of
1NE9.  1NE9 is a top hit for the whole 305 residue protein.  What
about 123-305?  Maybe we should run a make on the other portion of the
protein?

Wed Jul 19 13:48:27 PDT 2006 Chris Wong

Prof Karplus realized something in my .cshrc file was causing a bug in
make.  Fixed .cshrc file and redid the make2.  Running on vashon at
158p.


Wed Jul 19 14:54:26 PDT 2006 Chris Wong

The make 2 is done.  The make2 makes alignments, but not new models
from the aligments.  I'm going to set up try2.under to get the new
model.


( make -k T0372.do2 > & do2.log ; gzip -9f do2.log ) & started on
orcas at 300p.


Wed Jul 19 15:40:26 PDT 2006 Chris Wong

There's an error with known_ssbond.  Fixed and rerunning on orcas at
340p.

Thu Jul 20 14:56:57 PDT 2006 Kevin Karplus

The unconstrained cost fcn likes SAM_T06_server_TS1 and ROBETTA_TS4
better than either try1-opt2 or try2-opt2.  They should probably be
looked at.

Thu Jul 20 15:01:31 PDT 2006 Kevin Karplus

I'm wondering if the "intrasubunit disulfide bond" was supposed to be
an "intersubunit disulfide bond".  I hope not, but having a disulfide
bond a turn of a helix apart is a little unusual.

Fri Jul 21 12:01:24 PDT 2006 Chris Wong

Going to polish SAM_T06_server_TS1 as try3.  SAM_T06_server_TS1 looks
a lot like what we were getting with try1 and try2.  Started on lopez
at 1203p.

Fri Jul 21 16:37:17 PDT 2006 Chris Wong

try3-opt2 has many breaks.  Going to use the trick of polishing
T0372.try3-opt2.gromacs0.repack-nonPC.pdb to heal breaks.  I upped the
weight of breaks, too.  Try4 started on lopez at 1638.

Fri Jul 21 21:43:01 PDT 2006 Chris Wong

try4 healed some breaks, not all of them ,though.  It scores 3rd,
behind two models from try3.

Mon Jul 24 14:31:51 PDT 2006 Chris Wong

Going to do a try5 based on try4-opt2.  Adding some helix constraints
in the hopes that some nearby short helices will join.  Also, need to
heal the remaining breaks.  The helices will be the main focus,
though.

( make -k T0372.do5 > & do5.log ; gzip -9f do5.log ) & started on
orcas at 1444.

Wed Jul 26 13:40:08 PDT 2006 Chris Wong

try5 and try3 have similar unconstrained and constrained scores.  They
also look pretty much the same.  I'm going to increase the weight of
the constraints and breaks and run a try6.

( make -k T0372.do6 > & do6.log ; gzip -9f do6.log ) & started on
lopez at 1406.


Thu Jul 27 12:21:39 PDT 2006 Chris Wong

try6-opt2 has no breaks, but looks terrible.  I'm getting worried
about this one, and how to make it into a``dimeric oligomerization
state'' as the doc.html says.

So far, I've only worked on the SAM_T06_server_TS1 of the server
models (SAM_T06_server_TS1, try3, try4, try5 are in the same group).

I'm going to do a polishing run on the ROBETTA_TS4 model as try7.

( make -k T0372.do7 > & do7.log ; gzip -9f do7.log ) & started on
orcas at 1240p.


Thu Jul 27 19:48:24 PDT 2006 Chris Wong

Now I'm going to try to use try6.costfcn on the try7-opt2 model to see
if it does any better than what we got in try6.  This will be try8, in
the same group as ROBETTA_TS4 and try7.

( make -k T0372.do8 > & do8.log ; gzip -9f do8.log ) & started on
orcas at 752p.

Fri Jul 28 11:22:23 PDT 2006 Zack Sanborn

Well, try8 looks alright, but there are several "half-sheets" (beta
strands that look like they can be extended into neighboring strands)
that I will try to fix up with some intelligently chosen constraints.

I'm going to use the try8-opt2.sheets to do this, altering the
following strands:

SheetConstraint (T0372)F45 (T0372)I48 (T0372)F54 (T0372)F51 hbond
(T0372)I48 1 SheetConstraint (T0372)V53 (T0372)A58 (T0372)M66
(T0372)Q61 hbond (T0372)F54 1 SheetConstraint (T0372)T207 (T0372)N214
(T0372)M224 (T0372)I217 hbond (T0372)G208 1 SheetConstraint
(T0372)T221 (T0372)I226 (T0372)V235 (T0372)T230 hbond (T0372)F222 1
SheetConstraint (T0372)I226 (T0372)N227 (T0372)F231 (T0372)T230 hbond
(T0372)N227 1

And extend the first & fourth, strengthen the second & fourth, and
remove the final one

SheetConstraint (T0372)T43 (T0372)I48 (T0372)F56 (T0372)F51 hbond
(T0372)I48 1 SheetConstraint (T0372)V53 (T0372)A58 (T0372)M66
(T0372)Q61 hbond (T0372)F54 5 SheetConstraint (T0372)T207 (T0372)N214
(T0372)M224 (T0372)I217 hbond (T0372)G208 1 SheetConstraint
(T0372)A219 (T0372)I226 (T0372)K237 (T0372)T230 hbond (T0372)F222 4

I will also add these new constraints to build some strands where
there were predictions of beta sheets:

SheetConstraint (T0372)L62 (T0372)P67 (T0372)N90 (T0372)G95 hbond
(T0372)L62 1 SheetConstraint (T0372)F112 (T0372)D117 (T0372)N90
(T0372)G95 hbond (T0372)D117 1 SheetConstraint (T0372)E267 (T0372)Y261
(T0372)E236 (T0372)T230 hbond (T0372)Y261 1

I will also add the helix constraints from try8-opt2, and leave them
untouched.

try9, using the above constraints, and starting from
T0372.try8-opt2.pdb.gz, was started on orcas.

Fri Jul 28 13:13:55 PDT 2006 Zack Sanborn

Since we will need to dimerize whatever we get from these optimization
runs, I'd better prepare for it.

I'm going to attempt to dimerize from the 1mk4 template.  It is the
highest scoring alignment and is also a dimer!  Unfortunately, it is
only 158 residues, a little over half the size of T0372 monomer.  The
problem with this is that the alignment that we'll be basing the dimer
with might make a dimer that is poorly oriented.  We'll just have to
see, I guess.

The top-scoring t06 alignment that we'll use to make the dimer is
found here:
 
	1mk4A/T0372-1mk4A-t06-local-str2+near-backbone-11-0.8+0.6+0.8-adpstyle5.a2m

I've copied this over to the dimer directory and modified it in the
file 1mk4A.dimer-a2m.  I've tested the make-dimer script on try8-opt2
and it works.  Now, when we get a couple good templates, we might have
something.

Fri Jul 28 15:06:30 PDT 2006 Zack Sanborn

Well, try9 finished and has produced the highest-scoring model with an
unconstrained cost function.  But, it didn't do anything I wanted,
just shifted one of the "domains" a little w.r.t. the other.

I will try again, but will restart from the alignments this time.
That will have a better chance at making something I want, but at
greater risk of making something that looks terrible.


Fri Jul 28 16:31:08 PDT 2006 Zack Sanborn

I've made a dimer out of try9-opt2.  It makes a weird interface, but it's 
a reasonable dimer (no huge clashes, etc.).  For the fun of it, I want to
see how this dimer will be optimized, so I've started an optimization run
in the dimer directory

dimer/try1 started on whidbey.

( make -k T0372.mult1 > & mult1.log ; gzip -9f mult1.log ) &

I'm not sure if I should be worried, but I'm getting the following errors:

Error: Couldn't open file T0372.t2k.alpha.rdb or T0372.t2k.alpha.rdb.gz for input
Error: Couldn't open file T0372.t04.alpha.rdb or T0372.t04.alpha.rdb.gz for input
Error: Couldn't open file T0372.t06.alpha.rdb or T0372.t06.alpha.rdb.gz for input

It's still running, though, and I plan on just letting it run.  I'm not 
too worried since this error is for opening a file for "input" -- this 
usually means the file's missing.  I'm going to let it go and see what 
happens.


Sun Jul 30 15:51:25 PDT 2006 Zack Sanborn

Well, I'm glad I didn't worry.  The dimer was made and it made a pretty
tight interface.  However, the interface is very different than where
it started.  It started with the tips touching, and now the crevices 
of the dimers are positined together with one monomer rotated 90 degrees
relative to the other.  

Unfortunately, the interface is not a hydrophobic one when using the 
"near" script.  However, I don't really see many good places to line 
up the monomers.  Most have a patch of hydrophobics surrounded by
a lot of hydrophilics, or are equal scatterings of hydrophobic and 
hydrophilic residues. 

Well, I take that back a little bit.  It seems Undertaker chose the
"best" side of the protein to put together... it's just not a very 
good interface.  It might be better if the 90 degree rotation was
not there.  I'll try to ProteinShop a model like this and optimize
it with Undertaker.

Sun Jul 30 16:48:29 PDT 2006 Zack Sanborn

I've ProteinShop'd a model with the interfaces between the two monomers
like I *mostly* want them.  The results of this can be found in the 
dimer/decoys/T0372.model1_renum.pdb.gz.  I'm hoping Undertaker will 
take it the rest of the way for me.  Of course, it's very likely that
Undertaker will find the same interface it found for dimer/try1-opt2.

We'll just have to wait and see.

try2 started on camano.

Mon Jul 31 15:43:12 PDT 2006 Zack Sanborn

dimer/try2-opt2 is the top scoring model by just a little bit.  This
is a good thing, as it gives us a mostly different model (different
dimer, slightly different monomers) for the dimer submission.


Mon Jul 31 16:31:10 PDT 2006 Chris Wong

Now going to set up some polishing/optimizing runs.  try3 is for
try1-opt2 and try4 is for try2-opt2.

( make -k T0372.do3 > & do3.log ; gzip -9f do3.log ) & started on
vashon at 1635.

( make -k T0372.do4 > & do4.log ; gzip -9f do4.log ) & started on
chacha at 1641.


Tue Aug  1 10:53:51 PDT 2006 Zack Sanborn

Well, according to an unconstrained cost function, try4-opt2 is nearly
identical to try2-opt2, the model it was optimizing.  try3-opt2 showed
a *little* more improvement over try1-opt2, but still not by very much.

I don't know why such little improvement was shown, breaks, soft_clashes,
and dry weights were turned up pretty high.  Yet, the model changed
very little.  Curious.


Tue Aug  1 11:41:57 PDT 2006 Zack Sanborn

I'm going to make a chimera of the monomer models try7-opt2 < ROBETTA_TS4
and try9-opt2.  Both of these models are similar in their C-terminal domains,
but have different N-terminal domains.  This is expected given that 
the C-term domain has decent templates while the N-term domain does not.

The N-term domain will be taken from try7-opt2, residues D180-W305.  The
remaining residues (M1-D179) will be taken from our top-scoring model, 
try9-opt2.  This chimera will be optimized and turned into a dimer. 

Actually, scratch all that.  We don't really have time to do both things.
I will turn try7-opt2 into a dimer, and start an optimization of that.
The chimera will also be made, but will not likely be made into a dimer
due to time considerations.  The differences will likely be small.

The dimerization of try7-opt2 was done using the make-dimer2.under script,
based on the 1mk4 template like the other dimer.  The initial dimer
can be found in dimer-try2-1mk4A.pdb.

dimer/try5, optimizing dimer-try2-1mk4A.pdb, was started on orcas.

Tue Aug  1 12:00:11 PDT 2006 Zack Sanborn

The monomeric chimera was made of try9-opt2/M1-D179 and try7-opt2/D180-W305
and put into the file decoys/T0372.chimera.pdb.gz.  An optimization of
this chimera is try11.

try11 started on orcas.


Tue Aug  1 15:03:43 PDT 2006 Zack Sanborn

try11 has finished and scores pretty well.  It scores worse than the model
its based on, but I believe a polishing run will fix that.  It's not much
different than try7-opt2, where the N-terminal domain was taken, but does
score better due to fewer breaks/clashes. 

try12, a polishing run of try11-opt2, started on orcas.


Tue Aug  1 16:59:07 PDT 2006 Zack Sanborn

Well, it's time to attempt a submission of the dimers we have.  According
to an unconstrained cost function, the following models are the best:

	dimer/T0372.try4-opt2 < try2-opt2 < proteinshop < try1-opt2
	dimer/T0372.try2-opt2 < proteinshop to get different orientation of 
				  monomers < try1-opt2
  	dimer/T0372.try3-opt2 < try1-opt2
	dimer/T0372.try1-opt2 < dimer of try9-opt2 monomer using dimeric
 				  template 1mk4 < try8-opt2 < try7-opt2 
					< ROBETTA_TS4
	dimer/T0372.try5-opt2 < dimer of try7-opt2 < ROBETTA_TS4 

Unfortunately, these are all the models we have and it's unlikely we'll have
time to do any more.  I will update the superimpose-best.under script with these, 
then update the dimer/T0372.method file for submission.  


Tue Aug  1 17:32:02 PDT 2006 Zack Sanborn

Submitted the dimers, with comment:

All dimers were based on the PDB template 1mk4, our highest scoring template
from our automatic run.  Unfortunately, Undertaker had a hard time maintaining
the end-to-end orientation of the monomers found in template 1mk4.  However,
despite this, the interface between the monomers in the submitted dimeric models
are surprisingly tight.   

Model 1 is try4-opt2, the best scoring model with our unconstrained.costfcn
	cost function.  It was optimized by Undertaker from a model based on
	try2-opt2.  try2-opt2 was made using ProteinShop and our original 
	dimer try1-opt2. The monomers of try1-opt2 were rotated by 90 degrees
	to form a "flatter" dimer.  This orientation improved hydrophobic
	packing moderately.  The try1-opt2 dimer was made by taking our 
	monomer prediction, try9-opt2, and placing it according to the 
	orientation of the monomers in the dimeric PDB templated, 1mk4 
	using Undertaker.

Model 2 is try2-opt2, the second best scoring dimeric model using the 
	unconstrained.costfcn cost function.  As described above, it 
	originated from try1-opt2 after re-orientating the monomers using
	ProteinShop.  The ProteinShop model was optimized by Undertaker
	to reduce breaks and improve the packing.

Model 3 is try3-opt2, an optimization of try1-opt2, and our third highest
	scoring dimeric model.  This model has the monomers orientated in
	"cross" with a fairly tight interface at their center.  However, 
	large hydrophobic patches were left exposed, leading us to the 
	creation of a second dimer with the monomers situated in a better
	way. 

Model 4 is try1-opt2, which was a dimer made by taking our monomeric model
	try9-opt2 and dimerizing it according to the dimeric template 1mk4.
	This initial dimer was optimized by Undertaker to make try1-opt2.
	Interestingly, the start and final positions of these monomers 
	were very different.  Initially, the monomers were found end-to-end
	with a fairly weak interface.  After the optimization, the monomers
	were located in a cross-like formation, with the center of each 
	monomer providing the interface.

Model 5 is try5-opt2, the poorest scoring dimeric model among this group.
	It was made by dimerizing the monomer model try7-opt2, a direct
	optimization of the server model ROBETTA_TS4.  The N-terminal 
	domain of try7-opt2 differed significantly enough from try9-opt2
	to warrant a second dimeric model based on try7-opt2.  The dimer
	was again created by Undertaker using the dimeric template 1mk4.


Tue Aug  1 17:53:47 PDT 2006 Zack Sanborn

We should also prepare the monomer models for submission.  I know I don't have
to handle the submission this time, but I, of course, need to let whomever will
do the submission (Grant?) what models to submit, and I'll help them out by
updating the T0372.method file also. 


Tue Aug  1 18:43:07 PDT 2006 Zack Sanborn

I've updated the monomer T0372.method file to help out Kevin or Grant (or
whoever will be doing the monomer submissions).  I've added the following
to the T0372.method file:


The most successful models for T0372 were all based on the automatic server 
model ROBETTA_TS4.  This server model was judged by us to be the most 
likely out of any of our automatically generated models.  In fact, even our
automatic server prediction SAM_T06_server_TS1 was judged to be better 
than our initial automatic predictions.  Admittedly, more work should have
been performed on the SAM_T06_server_TS1 model instead of focusing all 
attention on ROBETTA_TS4 as we did.  

Model 1 is try12-opt2, our highest scoring model using the unconstrained
	cost function.  try12-opt2 was a polishing run of try11-opt2, which
	was an optimization of the chimera made from try9-opt2/M1-D179 and 
	try7-opt2/D180-W305.  This model is not too different from try7-opt2,
	but scores significantly better.

Model 2 is try9-opt2, our second highest scoring model using the 
	unconstrained cost function.  try9-opt2 was a polishing optimization 
	of try8-opt2 to reduce breaks and clashes and improve hydrophobic 
	packing.  try8-opt2 was itself an optimization of try7-opt2, a model
	based on an optimization of the automatic server model ROBETTA_TS4.

Model 3 is try11-opt2.gromacs0.repack-nonPC, which is the best scoring 
	model using the Rosetta cost function.  This model was made by 
	optimizing a chimera made from the C-terminal domain of try9-opt2
	(M1-D179) and the N-terminal domain of try7-opt2 (D180-W305).  The
	model was then re-optimized with gromacs then repacking the 
	sidechains (except for prolines and cysteines) by rosetta.

Model 4 is try7-opt2, which has the same C-terminal domain as try9-opt2, but
	has a different N-terminal domain.  This model was a direct 
	optimization of the automatic server model ROBETTA_TS4, the best 
	server model as judged by Undertaker's cost function.

Model 5 is try5-opt2, our highest scoring model that was not based on 
	the server model ROBETTA_TS4.  This model was made by optimizing
	try4-opt2 and trying to form some longer helices.  try4-opt2 
	came from try3-opt2, which resulted from an optimization of 
	our automatic server prediction SAM_T06_server_TS1.       

Tue Aug  1 19:54:52 PDT 2006 Kevin Karplus

Models submitted with comments as written by Zack---I saw nothing I
felt compelled to change.

You can continue to work on this for another day, if there is anything
you can think of doing in such a short time (especially given that 374
and 375 are also due in the same time frame).