Mon Aug 12 11:29:47 PDT 2002
t0180

No obvious homologies, short, so probably a good ab initio target.
Forms a dimer.  Appears to have a strand and 2 helices.
Lots of glycines may make it hard to model.
No homologs found by target2k script, so predictions are likely to be
poor. 

24 Aug 2002

How are we going to handle this?
Clearly we need to have the dimer if the strand is going to form a sheet.

We can pack the two helices so that their hydrophobic faces make contact.
We can force the strand to remain straight.  I don't know how we can
force the hydrophobic dimerization interface to appear.  I have a
conjecture how it looks---the two helices in an up-down bundle, with
the hydrophobic surface of the long helices making part of the
dimerization interface and the sheet (an anti-parallel connection)
almost at right angles to the pair of helices and sitting across them.

With the straightness constraint added and parameters tweaked to make
the hydrophobic surface a little less terrible, try1.5.60 now scores best.

Perhaps we should do another ab-initio run (no initial conformations
or alignments) and see what sorts of models come up.

25 Aug 2002 Kevin Karplus

best score now try2-opt---strand is still curling up.  Upping the
constraint weight to 0.5 makes a straight strand, but the helixes pack
poorly.  I'll do one more ab initio optimization with this revised
score function, and I'll also try creating a doubled chain, with two
copies of t0180 separated by a linker of 12 glycines.  I'll do this in
subdirectory t0180-double.

Fri Aug 30 16:45:53 PDT 2002

try3-opt is best with revised scoring function.
Helix S37-L50 seems to be oriented wrong, with its hydrophobic face out.

The t0180-double experiment did not work---it folded the two halves
wuite differently and did not respect the secondary structure.

For try4, let's try making a hairpin at the N-terminus.
There is a weakly predicted first strand.

// straightness constraints

constraint P15.CB R19.CB	12 14.6 16
constraint V16.CB L20.CB	12 14.6 16

constraint V2.CB R4.CB		6 7.3 8
constraint G3.CA R5.CA		6 7.3 8

// possible hairpin

constraint K11.CB	R5.CB	4 4.7 8
constraint D12.CB	R4.CB	4 4.7 8

With these constraints, try2-opt scores better than try3-opt.

Let's rerun from scratch.

31 Aug 2002 17:58	Kevin Karplus

Current best is try4-opt-scwrl

These actually look like they may pack into a dimer.
Without a program that actually handles dimers, I don't know that we
could do much better.

I'm going to submit this one and give up on it---maybe for CASP6 we'll
be able to handle multimers!

models:
1	try4-opt-scwrl
2	try2-opt
3	try3-opt