predict-2nd	
	part of the SAM-T02 suite

Tue Feb 19 15:42:15 PST 2008 Kevin Karplus

This source code is being released "as-is" for researchers who wish to
duplicate our work on protein local structure prediction.  Since we
have never received any funding for making an easily-installed,
user-friendly system, this releases is about all we can afford to do.

Since this directory is also our working directory for recent changes,
not a cleaned-up frozen snapshot of the code, it most likely contains
a lot of junk, as well as the essential parts of the code.

In addition to the code in predict-2nd, you will also need the
libraries in ../ultimate




Mon Jun 13 03:09:07 PDT 2005 Kevin Karplus

The following "Makefile" fragment, taken from the SAM-T02 server,
gives an example of running predict-2nd given an a2m file and an
already trained network.

############################################################
ROOTDIR=/projects/compbio/experiments/protein-predict/SAM_T02
HTMLDIR = /projects/compbio/.html/SAM_T02
WORKDIR= ${HTMLDIR}/results/${QUERYDIR}
BASEURL = http://www.soe.ucsc.edu/projects/compbio/SAM_T02/results/${QUERYDIR}
BIN = ${ROOTDIR}/bin_freeze
BIN-PREDICT2ND = ${BIN}/predict_2nd
PREDICT-2ND = ${BIN-PREDICT2ND}/predict-2nd
STR-NET = ${BIN-PREDICT2ND}/networks/t2k-5651-IDaaHr-5-15-7-15-9-15-13-str-seeded.net
%.t2k.str %.t2k.str.rdb %.t2k.str.seq : %.t2k-thin90.a2m.gz ${STR-NET}
	echo ReadAlphabet /projects/compbio/lib/alphabet/str.alphabet > tmp.script
	echo ReadNeuralNet ${STR-NET} >> tmp.script
	echo ReadA2M $< >> tmp.script
	echo PrintPrediction $*.t2k.str  >> tmp.script
	echo PrintPredictionFasta $*.t2k.str.seq >> tmp.script
	echo PrintRDB $*.t2k.str.rdb >> tmp.script
	${PREDICT-2ND} < tmp.script
	rm tmp.script
	${SCRIPTSDIR}/add_summary_html "STR structure prediction CASP format" ${BASEURL}/$*.t2k.str $*.t2k.str >> ${WORKDIR}/summary.html
	${SCRIPTSDIR}/add_summary_html "STR structure prediction RDB format" ${BASEURL}/$*.t2k.str.rdb $*.t2k.str.rdb >> ${WORKDIR}/summary.html
	${SCRIPTSDIR}/add_summary_html "STR EBGHTL structure prediction sequence format" ${BASEURL}/$*.t2k.str.seq $*.t2k.str.seq >> ${WORKDIR}/summary.html
############################################################

Currently, trained networks (and the scripts to train them) can be
found in /projects/compbio/usr/karplus/predict-2nd/testing/
roughly sorted into directories according to the type of alphabet
being predicted.

The alphabet files can be found in /projects/compbio/lib/alphabet/
Several alphabets may appear in the same file, which makes creating
<<<<<<< README
the make files needlessly difficult, since the mapping form alphabet
name to alphabet file is not obvious.  Changing this would require
making several changes in pipelines and makefiles that use
perdict-2nd, which we have not had time to do.

Sun Jun  1 09:04:32 PDT 2008 Kevin Karplus

alphabets/ subdirectory added to pre-release version on web, giving
access to all our alphabets.

Because std.alphabet location is hardwired in ultimate/src/Alphabet/Makefile,
that file needs to be changed before compiling on a different system.
If you need to run the program without recompiling, you can use
command-line option "-noalph" to skip reading in alphabets from
standard locations.  You will then need explicit ReadAlphabet commands
in your scripts, even for the standard alphabets.
=======
the make files needlessly difficult, since the mapping from alphabet
name to alphabet file is not obvious.
>>>>>>> 1.6



Thu Mar  5 16:04:02 PST 2009 Kevin Karplus

Note: the "zlib" package is needed by predict-2nd and ultimate, but is
not included here.  Look for it at
	http://www.zlib.net