filterBedByClassFreq

#!/bin/bash
#PBS -l nodes=1:ppn=4

MIN_N_CLASS=500
GENOME="mm9"
CLASS_COL=5

#### usage ####
usage() {
	echo Program: "filterBedByClassFreq (filter input BED file based on frequency of classes)"
	echo Author: BRIC, University of Copenhagen, Denmark
	echo Version: 1.0
	echo Contact: pundhir@binf.ku.dk
	echo "Usage: filterBedByClassFreq -i <file> [OPTIONS]"
	echo "Options:"
	echo " -i <file>   [input genomic coordinates in BED format (can be stdin)]"
    echo "             [chr start end stages_count class (..)]"
    echo "[OPTIONS]"
    echo " -n <int>    [minimum number of elements within class (default: 500)]"
    echo " -k <int>    [column number that contains stage information (default: 5)]"
	echo " -h          [help]"
	echo
	exit 0
}

#### parse options ####
while getopts i:n:k:h ARG; do
    case "$ARG" in
    i) BED_FILE=$OPTARG;;
    n) MIN_N_CLASS=$OPTARG;;
    k) CLASS_COL=$OPTARG;;
    h) HELP=1;;
    esac
done

## usage, if necessary file and directories are given/exist
if [ -z "$BED_FILE" -o "$HELP" ]; then
	usage
fi

## populating files based on input genome
if [ "$GENOME" == "mm9" ]; then
    GENOME_FILE="/home/pundhir/project/genome_annotations/mouse.mm9.genome"
elif [ "$GENOME" == "hg19" ]; then
    GENOME_FILE="/home/pundhir/project/genome_annotations/human.hg19.genome"
else
    echo "Presently the program only support analysis for mm9 or hg19"
    echo
    usage
fi

## create temporary BED file if input is from stdin
if [ "$BED_FILE" == "stdin" ]; then
    TMP=$(cat /dev/urandom | tr -dc 'a-zA-Z0-9' | fold -w 32 | head -n 1)
    while read LINE; do
        echo ${LINE}
    done | perl -ane '$line=""; foreach(@F) { $line.="$_\t"; } $line=~s/\t$//g; print "$line\n";' > $TMP
    BED_FILE=$TMP
fi

## filter input BED file based on minimum number of elements within each nfr dynamic class
for CLASS in $(cat $BED_FILE | cut -f $CLASS_COL | sort | uniq -c | sed -E 's/^\s+//g' | sort -k 1rn,1 | perl -ane 'if($F[0]>'$MIN_N_CLASS') { print "$F[1]\n"; }'); do perl -ane 'if($F['$((CLASS_COL-1))']=~/^'$CLASS'$/) { $F['$((CLASS_COL-1))']=~s/\,/\_/g; print "$F[0]"; foreach(@F[1..scalar(@F)-1]) { print "\t$_"; } print "\n"; }' $BED_FILE; done | sort -k 1,1 -k 2n,2 -k 3n,3

## remove temporary file, if exists
if [ ! -z "$TMP" ]; then
    rm $TMP
fi