comparison tools/encode/random_intervals_no_bits.py @ 0:9071e359b9a3

Uploaded

0:9071e359b9a3

#!/usr/bin/env python
#Dan Blankenberg
#%prog bounding_region_file mask_intervals_file intervals_to_mimic_file out_file mask_chr mask_start mask_end interval_chr interval_start interval_end interval_strand use_mask allow_strand_overlaps
import sys, random
from copy import deepcopy
from galaxy import eggs
import pkg_resources
pkg_resources.require( "bx-python" )
import bx.intervals.io
import bx.intervals.intersection
import psyco_full

assert sys.version_info[:2] >= ( 2, 4 )

max_iters = 5

def stop_err( msg ):
    sys.stderr.write( msg )
    sys.exit()

#Try to add a random region
def add_random_region( mimic_region, bound, exist_regions, plus_mask, minus_mask, overlaps ):
    region_length, region_strand = mimic_region
    plus_count = plus_mask.count_range()
    minus_count = minus_mask.count_range()
    gaps = []

    if region_strand == "-":
        gaps = minus_mask.get_gaps( region_length )
    else:
        gaps = plus_mask.get_gaps( region_length )
    
    while True:
        try:
            gap_length, gap_start, gap_end = gaps.pop( random.randint( 0, len( gaps ) - 1 ) )
        except:
            break
        try:
            start = random.randint( bound.start + gap_start, bound.start + gap_end - region_length - 1 )
        except ValueError, ve:
            stop_err( "Exception thrown generating random start value: %s" %str( ve ) )

        end = start + region_length
        try_plus_mask = plus_mask.copy()
        try_minus_mask = minus_mask.copy()
        
        if region_strand == "-":
            try_minus_mask.set_range( start - bound.start, end - bound.start )
        else:
            try_plus_mask.set_range( start - bound.start, end - bound.start )
        
        rand_region = bx.intervals.io.GenomicInterval( None, [bound.chrom, start, end, region_strand], 0, 1, 2, 3, "+", fix_strand=True )
        
        if try_plus_mask.count_range() == plus_count + region_length or try_minus_mask.count_range() == minus_count + region_length:
            if overlaps in ["strand", "all"]: #overlaps allowed across strands
                exist_regions.append( rand_region )
                if overlaps == "strand":
                    return exist_regions, True, try_plus_mask, try_minus_mask
                else: #overlaps allowed everywhere
                    return exist_regions, True, plus_mask, minus_mask
            else: #no overlapping anywhere
                exist_regions.append( rand_region )
                if region_strand == "-":
                    return exist_regions, True, try_minus_mask.copy(), try_minus_mask
                else: 
                    return exist_regions, True, try_plus_mask, try_plus_mask.copy()
    return exist_regions, False, plus_mask, minus_mask

def main():
    includes_strand = False
    region_uid = sys.argv[1]
    mask_fname = sys.argv[2]
    intervals_fname = sys.argv[3]
    out_fname = sys.argv[4]
    try:
        mask_chr = int( sys.argv[5] ) - 1
    except:
        stop_err( "'%s' is an invalid chrom column for 'Intervals to Mask' dataset, click the pencil icon in the history item to edit column settings." % str( sys.argv[5] ) )
    try:
        mask_start = int( sys.argv[6] ) - 1
    except:
        stop_err( "'%s' is an invalid start column for 'Intervals to Mask' dataset, click the pencil icon in the history item to edit column settings." % str( sys.argv[6] ) )
    try:
        mask_end = int( sys.argv[7] ) - 1
    except:
        stop_err( "'%s' is an invalid end column for 'Intervals to Mask' dataset, click the pencil icon in the history item to edit column settings." % str( sys.argv[7] ) )
    try:
        interval_chr = int( sys.argv[8] ) - 1
    except:
        stop_err( "'%s' is an invalid chrom column for 'File to Mimick' dataset, click the pencil icon in the history item to edit column settings." % str( sys.argv[8] ) )
    try:
        interval_start = int( sys.argv[9] ) - 1
    except:
        stop_err( "'%s' is an invalid start column for 'File to Mimick' dataset, click the pencil icon in the history item to edit column settings." % str( sys.argv[9] ) )
    try:
        interval_end = int( sys.argv[10] ) - 1
    except:
        stop_err( "'%s' is an invalid end column for 'File to Mimick' dataset, click the pencil icon in the history item to edit column settings." % str( sys.argv[10] ) )
    try:
        interval_strand = int( sys.argv[11] ) - 1
        includes_strand = True
    except:
        interval_strand = -1
    if includes_strand:
        use_mask = sys.argv[12]
        overlaps = sys.argv[13]
    else:
        use_mask = sys.argv[11]
        overlaps = sys.argv[12]
    available_regions = {}
    loc_file = "%s/regions.loc" % sys.argv[-1]
    
    for i, line in enumerate( file( loc_file ) ):
        line = line.rstrip( '\r\n' )
        if line and not line.startswith( '#' ):
            fields = line.split( '\t' )
            #read each line, if not enough fields, go to next line
            try:
                build = fields[0]
                uid = fields[1]
                description = fields[2]
                filepath = fields[3]
                available_regions[uid] = filepath
            except:
                continue

    if region_uid not in available_regions:
        stop_err( "Region '%s' is invalid." % region_uid )
    region_fname = available_regions[region_uid].strip()

    #set up bounding regions to hold random intervals
    bounds = []
    for bound in bx.intervals.io.NiceReaderWrapper( open( region_fname, 'r' ), chrom_col=0, start_col=1, end_col=2, fix_strand=True, return_header=False, return_comments=False ):
        bounds.append( bound )
    #set up length and number of regions to mimic
    regions = [ [] for i in range( len( bounds ) ) ]

    for region in bx.intervals.io.NiceReaderWrapper( open( intervals_fname, 'r' ), chrom_col=interval_chr, start_col=interval_start, end_col=interval_end, strand_col=interval_strand, fix_strand=True, return_header=False, return_comments=False ):
        #loop through bounds, find first proper bounds then add
        #if an interval crosses bounds, it will be added to the first bound
        for i in range( len( bounds ) ):
            if bounds[i].chrom != region.chrom:
                continue
            intersecter = bx.intervals.intersection.Intersecter()
            intersecter.add_interval( bounds[i] )
            if len( intersecter.find( region.start, region.end ) ) > 0:
                regions[i].append( ( region.end - region.start, region.strand ) ) #add region to proper bound and go to next region
                break
    for region in regions:
        region.sort()
        region.reverse()
    
    #read mask file
    mask = []
    if use_mask != "no_mask":
        for region in bx.intervals.io.NiceReaderWrapper( open( mask_fname, 'r' ), chrom_col=mask_chr, start_col=mask_start, end_col=mask_end, fix_strand=True, return_header=False, return_comments=False ):
            mask.append( region )

    try:
        out_file = open ( out_fname, "w" )
    except:
        stop_err( "Error opening output file '%s'." % out_fname )

    i = 0
    i_iters = 0
    region_count = 0
    best_regions = []
    num_fail = 0
    while i < len( bounds ):
        i_iters += 1
        #order regions to mimic
        regions_to_mimic = regions[i][0:]
        if len( regions_to_mimic ) < 1: #if no regions to mimic, skip
            i += 1
            i_iters = 0
            continue 
        #set up region mask
        plus_mask = Region( bounds[i].end - bounds[i].start )
        for region in mask:
            if region.chrom != bounds[i].chrom: continue
            mask_start = region.start - bounds[i].start
            mask_end = region.end - bounds[i].start
            if mask_start >= 0 and mask_end > 0:
                plus_mask.set_range( mask_start, mask_end )
        minus_mask = plus_mask.copy()
        random_regions = []
        num_added = 0
        for j in range( len( regions[i] ) ):
            random_regions, added, plus_mask, minus_mask = add_random_region( regions_to_mimic[j], bounds[i], random_regions, plus_mask, minus_mask, overlaps )
            if added: 
                num_added += 1
        if num_added == len( regions_to_mimic ) or i_iters >= max_iters:
            if len( best_regions ) > len( random_regions ):
                random_regions = best_regions.copy()
            num_fail += ( len( regions_to_mimic ) - len( random_regions ) )
            i_iters = 0
            best_regions = []
            for region in random_regions:
                print >>out_file, "%s\t%d\t%d\t%s\t%s\t%s" % ( region.chrom, region.start, region.end, "region_" + str( region_count ), "0", region.strand )
                region_count += 1
        else:
            i -= 1
            if len( best_regions ) < len( random_regions ):
                best_regions = random_regions[:]
        i+=1
    
    out_file.close()
    if num_fail:
        print "After %i iterations, %i regions could not be added." % (max_iters, num_fail)
        if use_mask == "use_mask":
            print "The mask you have provided may be too restrictive."

class Region( list ):
    """
    A list for on/off regions
    """
    def __init__( self, size=0 ):
        for i in range( size ):
            self.append( False )
    def copy( self ):
        return deepcopy( self )
    def set_range( self, start=0, end=None ):
        if start < 0:
            start = 0
        if ( not end and end != 0 ) or end > len( self ):
            end = len( self )
        for i in range( start, end ):
            self[i]=True
    def count_range( self, start=0, end=None ):
        if start < 0:
            start = 0
        if ( not end and end != 0 ) or end > len( self ):
            end = len( self )
        return self[start:end].count( True )
    def get_gaps( self, min_size = 0 ):
        gaps = []
        start = end = 0
        while True:
            try: 
                start = self[end:].index( False ) + end
            except: 
                break
            try: 
                end = self[start:].index( True ) + start
            except:
                end = len( self )
            if end > start and end - start >= min_size:
                gaps.append( ( end - start, start, end ) )
        gaps.sort()
        gaps.reverse()
        return gaps

if __name__ == "__main__": main()