07 May 2009 Grant Thiltgen

In each of the alphabet directories (bys, str, burial...), there is a small
Makefile that sets a few alphabet specific-variables and then does an
'include' of Make.alphabet in the parent directory.

If you create a new alphabet directory (or update an existing one) copy the
template Makefile from starter-directory and edit it as appropriate.

Then use the 'make' targets in Make.alphabet to create scripts or other
files as needed, by invoking 'make' in the alphabet directory, possibly
with various other make macros set on the command line.

In order to create the subdirectories needed for each new alphabet run:
	make dirs

After finishing the prerequisites listed below, and you are ready for the 
testing, there are many makes to be run.  The makes must be run in the 
following order, and each one must be finished before running the next one:
	make trAB-mult50.scripts
	make trAB-mult50.run
	make collect-mult50-data
	make trAB-mult150.scripts
	make trAB-mult150.run
	make collect-mult150-data
	make trAB-mult250.scripts
	make trAB-mult250.run
	make collect-mult250-data
	make trAB-test.scripts
	make trAB-test.run
	make collect-test-data

The final report will be in the quality-reports subdirectory.

Sat Jul  9 22:38:48 PDT 2005 Kevin Karplus

Some prerequisites:

In the networks subdirectory of the alphabet directory, there must be
a blank network of the appropriate structure (currently
	IDGaaH13-3-13-7-13-9-13-11-pb-empty.net
	
In the training-data subdirectory of this directory, there must be the
3-fold cross-validation set based on dunbrack-30pc-1763.
For example, the following commands make the t04 cross-validation set
for the t04-thin90 alignments:
make dunbrack-30pc-1763-1.t04-thin90-pb  dunbrack-30pc-1763-2.t04-thin90-pb  dunbrack-30pc-1763-3.t04-thin90-pb


In the pce/models.97/pdb/ directories for the chains of the training
and test set, there must be the appropraite correct local structure
file: for example
	4u/4ubpA/info/4ubpA.stride-mixed.pb

In pc/lib/alphabet/dunbrack-in-scop.backgrounds, make sure that there
is a set of background probabilities for the alphabet being trained for.
Probabilities vectors can be found for most alphabets in
~karplus/dna/predict-2nd/compare-real/comparisons/ 
where there are contingency tables comparing different alphabets.

Everything listed below is the old protocol and does not need to be run.

Sun Jul 10 09:52:34 PDT 2005 Kevin Karplus

Read the instructions in starter-directory/README


When you are ready, do
	make TXXALIGN=t04-thin90 tAcB.scripts run-tAcB
	make TXXALIGN=t2k-thin90 tAcB.scripts run-tAcB
Note: for the paper, we'll need to compare 2 architectures also--see starter-directory/run_makescripts:
	make TXXALIGN=t04-thin90 ARCHITECTURE=IDaaHr-5-15-7-15-9-15-13 tAcB.scripts run-tAcB
	make TXXALIGN=t2k-thin90 ARCHITECTURE=IDaaHr-5-15-7-15-9-15-13 tAcB.scripts run-tAcB

or use the "para create" mechanism described in starter-directory/README