Tue Aug 13 12:45:48 PDT 2002
t0182

31 August 2002 Kevin Karplus

Obvious homology model for 1c24A.

try1-opt-scwrl looks pretty good.
One moderate break at H231-F232---looks easily fixed.

No ssbonds or cys clusters for metal binding.

Upping dryness and break weights and adding hydrogen_bonds to scoring
function makes try1.13.60 score better.

I'll try two polishing runs
	try2	starts from best few models and uses hydrogen bonding info
	try3	starts from best few models and doesn't use hydrogen
			bonding info

19:10
OOPS forgot to include scwrl_each and use_all in OptConform---let's
kill these and try again.  I might as well wait until first iterations
are done and include those in the subsequent runs---that should take
less than 20 minutes more.

condor jobs to kill are 55499 and 55500.


20:21 started try4 without hbonds, and try5 with hbonds.

Without hbonds, current best is try2.0.120, then try3.1.120.
With hbonds, first few are ordered the same.

1 Sept 2002 Kevin Karplus 11:28

The run with Hbonds is meaningless, because bugs in the programs
caused other parts of the score function to be miscomputed.


The best score currently is try4-opt.
There are a few small breaks that might be fixed with a little
jiggling and CloseGap.

If I twiddle the parameters a bit, including increasing the break
weight, try4.3.120 scores best.  Let's do another (shorter) polishing
run from the best few models.


17:49  

try6-opt is the new best.
The remaining breaks are tiny (the biggest one is a stretched
peptide plane at F160-N164).

Let's submit it.