0. needs the latest version of ERANGE and the SOMCODE:
export ERANGEPATH=/woldlab/loxcyc/data00/alim/commoncode
export SOMPATH=/woldlab/loxcyc/data00/alim/testcode/SOM

Also make sure that there is a working version of the som c-extension 
compiled and copied into the SOM directory. If necessary, you can 
recompile the extension using the following commands:

cd $SOMPATH
rm -r $SOMPATH/build
python extension_setup.py build
cp $SOMPATH/build/lib*/_csom.so .

1. create a file that lists the ERANGE regions to use for partitioning the 
genome of the form:
/path/to/data/factor1.regionsLC.txt
/path/to/data/factor2.regionsLC.txt
...

You can skip this step if you only want to train the SOM on a finite set of 
regions (as long as they are in a single region file).

2. create a file that contains the datasets to be scored on the partition
of the form:
someid1,/path/to/data/factor1.rds
someid2,/path/to/data/factor2.rds
....

3. run the partition.sh script with a name, a minimum partition size, and 
the region files (resulting file myotubechromN.part where N is the number 
of region files).
. $ERANGEPATH/partition.sh myotubechrom 400 ../partition.files

4. get the RPKMs for each of the dataset in the dataset.files file.
. $ERANGEPATH/regionCounts.sh myotubechrom6.part ../chrommyotube.files 3000000

5. build a matrix of the dataset (resulting file: myotubechrom.matrix.tab).
. $ERANGEPATH/buildMatrix.sh myotubechrom ../chrommyotube.files 100 -rescale

6. build the SOM. This only save the weights of the SOM.
python $SOMPATH/trainsom.py 40 60 myotubechrom.matrix.tab myotubechrom40_60.som -trials 3 -timesteps 4000000

7. score the SOM. Scores the SOM on a dataset (in this case, the dataset used 
for training).
python $SOMPATH/scoresom.py myotubechrom40_60.som myotubechrom.matrix.tab myotubechrom40_60.scores

8. Map the SOM. Save the region distribution, component plane weights, and 
various dataset maps and PNGs in the somvis2 directory with the myo_all prefix.
python $SOMPATH/mapsom.py myotubechrom40_60.som myotubechrom40_60.scores somvis2/myo_all -data myotubechrom.matrix.tab -all -dataheader -savemap

Note that you can actually do a translation of the SOM (and hence the 
PNGs/maps) by dx,dy using the additional parameter -translate, e.g. 
-translate -15,15

9. Subtracting two maps to get a third map (and PNG). Be careful to only mix
maps that are from the same SOM and have been translated to the same set of 
coordinates.
python $SOMPATH/diffmap.py ../myotubechrom40_60.som ../myotubechrom40_60.scores NRSF_60h_10686.map Control_60h_10194.map NRSF_60h-control.map -png NRSF-control -title 'NRSF 60h - control'

10. Can build additional matrices using steps 4 & 5, which can then be mapped 
using step 9 *without retraining or rescoring the SOM*, by just changing the 
-data <matrix> parameter or mapsom.py.

11. Another way to generate map differences quickly is to create a 3-column, 
tab-delimited file (named combined.diffs in the example below) listing the 
newmap name,starting map, and map to subtract, using the matrix column names,
e.g.:
Myog_diff   Myog_60h_10158  Myog_exp_10599
Myog_rdiff  Myog_exp_10599  Myog_60h_10158
Myog_60h_clean  Myog_60h_10158  Control_60h_10194
Myog_exp_clean  Myog_exp_10599  Control_exp_10135
Myog_diff_clean Myog_60h_clean  Myog_exp_clean

Note that you can reuse computed difference maps in later maps as long as they 
were calculated earlier.

You can then use this file with mapsom.py to generate those maps at the same 
time as the ones in your scoring matrix:

python $SOMPATH/mapsom.py myoboth200.som myoboth200.scores 200som/combScaled -data combScaled.matrix.tab -all -dataheader -translate -20,10 -savemap -diffs combined.diffs -count

In this case, the additional -count flag will score the sum of the signal in the unit, rather than the average.