Fri Aug 13 16:57:23 PDT 2004
T0278


Fri Aug 13 21:41:39 PDT 2004 Kevin Karplus

comparative model with 1tqnA and many other a.104.1.1 templates.


Sat Aug 14 13:02:18 PDT 2004 Kevin Karplus

The multiple alignments are large (3000 sequences for t04, 4000 for
t2k), which means that the mutual information computation is slow, but
there are a lot of strong MI signals, even when thinning to 30% (since
t2k still has 166 sequences at 30% id).

I wonder whether the cytochromes will survive undertaker optimization,
since we don't have the heme groups represented.

The try1-opt2 model looks pretty good.  I think I'll just polish it
with an unconstrained cost function and call it done.


Sun Aug 15 15:50:04 PDT 2004 Kevin Karplus

OK, try2-opt2 looks fine also.

I'll submit
	try2-opt2	best undertaker
	try1-opt2	full auto
	try1-opt2.repack-nonPC	best Rosetta energy
	T0278-1tqnA-t2k-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5	best model
	T0278-1nr6A-t04-local-str2+CB_burial_14_7-0.4+0.4-adpstyle5 	2nd best with t2k


Sat Sep 25 12:23:12 PDT 2004 Kevin Karplus

We really nailed this target with 94-97% of smooth GDT.  That is
hardly surprising though, since the correct structure (1n97A) was in
the PDB before we were given the target.

The template-library alignments only got about 50% GDT, as did
Robetta, since the template library had not been updated recently enough.