Thu Dec 30 11:38:00 PST 2010 Kevin Karplus

phizetagauss7_hbond_21 had a high mutual information with amino acid,
so I'm testing it for use as a design alphabet.  I expect it to do
slightly worse than r_rot_psi_omega_16_mod, which had a higher mutual
info. 

First I made sure that ~/pcem/Makefile.models97 had the alphabet in 
the definition of BASIC_SETUP_TARGETS:		
		info/${CHAIN}.stride-mixed.phizetagauss7_hbond_21 \
Then I did a "serial-make basic-setup" with the
dunbrack-30-pc-1763.ids  and dunbrack-in-scop.ids as input. 

In ../training-data/ I made
make dunbrack-30pc-1763-1.phizetagauss7_hbond_21-aa dunbrack-30pc-1763-2.phizetagauss7_hbond_21-aa dunbrack-30pc-1763-3.phizetagauss7_hbond_21-aa

I copied ../pb_rev/Makefile here and modified it to have the correct
alphabet names and alphabet file.

Then I ran (in ./)
	make dirs
to create the directories, and created
	networks/Gphizetagauss7_hbond_21-5-20-5-20-5-aa-empty.net 
based on the similar network in ../pb_rev/networks/

Then I ran 
	make trAB-mult50 to start the training.
	
Fri Dec 31 07:58:41 PST 2010 Kevin Karplus

bit gain (from quality-reports/)			
epochs	arch			tr12avg	tr12best	tr23avg	tr23best	tr31avg	tr31best
phizetagauss7_hbond_21
50	5-20-5-20-5		0.17657	0.2207		0.17436	0.2253		0.17567	0.2229

Fri Dec 31 13:19:51 PST 2010 Kevin Karplus

150	5-20-5-20-5		0.29312	0.3034		0.29284	0.3040		0.29593	0.3082

Fri Dec 31 18:49:14 PST 2010 Kevin Karplus

250	5-20-5-20-5		0.33133	0.3329		0.33167	0.3350		0.3323	0.3366
test	5-20-5-20-5			0.3244			0.3284			0.3328

Sequence recovery
test	5-20-5-20-5			18.71%			18.79%			18.85%