Thu Aug  8 21:53:10 PDT 2002
t0175

9 Aug 2002

Homology model  1kpgA c.66.1.18

10 Aug 2002

Try1-opt-scwrl looks pretty good, but one strand has detached itself
from the sheet.
The hbonding positions on the exposed strand are M136, I138, and E140.
It is a little difficult to see what they should hbond to in
try1-opt-scwrl, but the detached strand is antiparallel, approximately
D180-L186.
	134	GIMLIGE
	180	DVVEMVL

Interestingly, these strands are NOT hbonded in 1kpgA.
That structure has 5 parallel strands, then an antiparallel pair.
	^^^^v^
The one antiparallel strand is the C-terminal end of 1kpgA, and is not
aligned in most of the 1kpgA alignments to T0175.

We are definitely predicting 6 strands, not 7, and D180-L186 is
predicted to be antiparallel edge, so perhaps we should be aligning it
to the end of 1kpgA instead of to the previous strand.

I'll have to hand-edit an alignment to do this, but I have no idea
where to put the helices that come after 190---perhaps they replace
the helices that come between the last two strands in 1kpgA?
Maybe I should look at some of the other possible templates also, to
see if there is one that has 6 strands ^^^^^v instead of 7 ^^^^v^.

24 Aug 2002 Kevin Karplus

Since I'm too overwhelmed with targets to do much thinking, I'll take
an easy out and just try to attach the strand in parallel:

M136.O	E183.N
I138.N	E183.O
I138.O  V185.N	

I'll add these constraints as try2.constraints.
try1-opt-scwrl is still best scorer after adding these.

25 Aug 2002 Kevin Karplus

try2-opt looks pretty good, but we need to extend the hbond
constraints back further on the strand we are trying to attach.

29 Aug 2002 22:48 Kevin Karplus

try3-opt scores best with try3.constraints and with no constraints.
There are a few small breaks and clashes that might be worth doing a
polishing run to clean up (doing try4 with no constraints, try5 with
the try3 constraints).  Hmm, I forgot to turn the break and clash
weights up for try4---I'll do it for try5 though.

Fri Aug 30 14:30:20 PDT 2002 Kevin Karplus

try5-opt scores best.  There is a bit of a break at the N terminal
ends of the strands, but it's not bad.


The C-terminal helix should bend at P211 and P227, but bends slightly
elsewhere Let's turn up break and pred_alpha2 and reoptimize.
With them turned up, try4-opt scores better.

22:18

try6-opt is new best.  I doubt we'll get significant improvements.
Let's submit this.