ridassaf / islands

#####################################################################################
First, all island headers are written to 'sorted-allislands.headers'

Then, run:

python generateNegatives.py sorted-allislands.headers

to get 'negatives.headers' as an output file with random coordinates. 

then run:

python repeatFeatures.py negtives.headers 

to get 'features'

I have created multiple versions of those named negative-features-x

To test, run 

python classify.py positive-features negative-features-x file1 file2 N 

file1 and file2 can be any feature files (my way of creating different versions
to train and test on different data, for these experiments they can be disregarded). 
N is the number of lines in file2. 
#####################################################################################
Another way is after getting the positive and the negative features using 
repeatFeatures.py, cut the Start and End away, then use:
cat features | perl -pe '$_="1\t$_"' > labeled-features
to label the features, then cat the header followed by the features files to 
create a combined feature file(many of those in the folder), then use:

python classifang.py combined-features -s 0 -t 0.5 -f 5

to classify using different machine learning algorithms. F stands for fold, s
is sensitivity, I forgot what t is. 

This also saves a random forrest model that can be loaded later to test.
#####################################################################################
singleWindowClassify.py takes a single window features file and classifies. 
It uses the model previously saved by classifang. 

filterRepeats.py filters repeat in two levels. 
attLORFinder.py gets repeats before being filtered. An update to repeatFinder.py
repeatSurfer.py was a logical attempt to get islands out of repeats.
#####################################################################################
In Reference Genomes:
allIslands.headers: contains all islands curated by surfer/mislander. 
allIslandsHeaders.sh: is the script that got them. 
count.sh and compare.sh are scripts to get number of islands and compare them.
download.sh is a script to download genomes given ids. 
ReferenceGenomes.ids is a file with the reference genomes ids. 
generateNegatives.py uses sorted-allIslands.headers to generate false islands.
singleWindowMislanderFeatures: uses negative.headers to generate single window
	features.
singleWindowRepeatFeatures: uses allIslands.headers to generate single window
	features.
#####################################################################################