Mon Aug  9 15:40:38 PDT 2004
T0269

DUE 29 Aug 2004


Tue Aug 10 01:49:30 PDT 2004 Kevin Karplus

comparative model, 1qmvA (c.47.1.10) as template.

There are a couple of C-terminal strands that need to be attached to
the sheet.  If they don't attach automatically in try2, I'll have to
try adding some sheet constraints.

Unfortunately, we don't seem to have any rr constraints for residues
after K168. (well, very weak ones for {154,128}-187, 113-196,
{73,111,112}-200).


Tue Aug 10 12:19:06 PDT 2004 Kevin Karplus

I like try2-opt2 a lot better than try1-opt2---the C-terminal domain
seems much more convincing.  Although we could break it off, optimize
it separately, and reattach it, the effort does not seem worth it
right now.

Oh, why not? the coverage of the multiple alignment is pretty weak
after M140, but the structures in t2k.undertaker-align look good
through G164, so let's start with D165.


Tue Aug 10 14:10:52 PDT 2004 Kevin Karplus

The 165-end domain is coming up with 1prxA as the top hit for both t2k
and t04 methods, though the methods don't agree much past that, and
neither one has many sequences as homologs.  The t2k alignment looks a
little deeper than the t04 one for the subdomain, so we'll rely more
on that.  The alignments in the t2k.undertaker-align.pdb file are all
quite short, so this subdomain prediction is going to be mostly ab initio.
I'll make the extra_alignments for the subdomain, so that the second
run can have more starting points to work from.


Tue Aug 10 14:43:20 PDT 2004 Kevin Karplus

Oops---there seems to be a problem.  I forgot to put the
START_COL:=165 line in the Makefile for the subdirectory.
I'll have to redo the setup, then run try2.


Tue Aug 10 15:04:34 PDT 2004 Kevin Karplus

Foo, more tired that I realized---I added the START_COL:= to the wrong
Makefile.  Backing up and trying again!


Wed Aug 11 15:37:13 PDT 2004 Kevin Karplus

I'm now up to try3 on the 165-end subdirectory, and it is looking
fairly good.  I see no reason to do further optimization of the
subdomain as an independent unit---now it is time to paste it in and
try to optimize the whole thing.

I wrote make-chimera.under to do the superposition, then did
cut-and-paste to create decoys/try2-chimera.pdb.

The chimera has bad breaks and clashes, but some repositioning should
fix that quickly.


Wed Aug 11 21:10:29 PDT 2004 Kevin Karplus


Foo! try3.under still had the try2 names, so overwrote the old try2 output!
Now I can't tell whether the old try2 or the new try3 models are better!
I've renamed the try3 outputs to use try3, to reduce confusion, but 
I didn't have the try2 outputs on my laptop even, so they are gone,
gone, gone.

Wait---I can get back try2-opt2.pdb from the
try2+165-end-try3-models.pdb file.

oops, no need---the try2 files were still there gzipped.
The try2 models still score slightly better than the try3 one, with
the try3 costfcn.

I wonder what I should try next?  reoptimizing try2 and try3 together?
adding inter-domain constraints? removing constraints?


Sun Aug 15 10:35:59 PDT 2004 Kevin Karplus

I'm just going to do a polishing run with an unconstrained costfcn.


Sun Aug 15 17:19:04 PDT 2004 Kevin Karplus

try4-opt2 looks pretty good.
I notice that C213 and C207 come very close---they may need to form a
disulfide, since this is a thioredoxin.  What isn't clear is whether
the crystal will have the oxidized or reduced form of the protien.
Still--it is probalby better to have the disulfide formed than to have
it impossible to form.

I'll do another polishing run with an SSBond constraint.


Mon Aug 16 07:37:05 PDT 2004 Kevin Karplus

Although there is a little looseness around A234, I think try5-opt2 is
about as good as we'll get in the time we have.

I'll submit 
	try5-opt2	best with disulfide
	try4-opt2	best without disulfide
	try5-opt2.repack-nonPC	best Rosetta energy
	try1-opt2	full auto
	T0269-1qmvA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5