def main():
region_fname = sys.argv[1]
mask_fname = sys.argv[2]
nsamples = int( sys.argv[3] )
intervals1_fname = sys.argv[4]
intervals2_fnames = sys.argv[5:]
nfeatures = len( intervals2_fnames )
total_actual = zeros( nfeatures )
# total_lengths1 = 0
total_lengths2 = zeros( nfeatures )
total_samples = zeros( ( nsamples, nfeatures ) )
for line in open( region_fname ):
# Load lengths for all intervals overlapping region
fields = line.split()
print >>sys.stderr, "Processing region:", fields[3]
r_chr, r_start, r_stop = fields[0], int( fields[1] ), int( fields[2] )
r_length = r_stop - r_start
# Load the mask
mask = overlapping_in_bed( mask_fname, r_chr, r_start, r_stop )
bits_mask = as_bits( r_start, r_length, mask )
bits_not_masked = bit_clone( bits_mask ); bits_not_masked.invert()
# Load the first set
intervals1 = overlapping_in_bed( intervals1_fname, r_chr, r_start, r_stop )
bits1 = as_bits( r_start, r_length, intervals1 )
# Intersect it with the mask
bits1.iand( bits_not_masked )
# Sanity checks
assert count_overlap( bits1, bits_mask ) == 0
# For each data set
for featnum, intervals2_fname in enumerate( intervals2_fnames ):
print >>sys.stderr, intervals2_fname
intervals2 = overlapping_in_bed( intervals2_fname, r_chr, r_start, r_stop )
bits2 = as_bits( r_start, r_length, intervals2 )
bits2.iand( bits_not_masked )
assert count_overlap( bits2, bits_mask ) == 0
# Observed values
actual_overlap = count_overlap( bits1, bits2 )
total_actual[featnum] += actual_overlap
# Sample
lengths2 = list( interval_lengths( bits2 ) )
total_lengths2[ featnum ] += sum( lengths2 )
for i in range( nsamples ):
# Build randomly covered bitmask for second set
random2 = throw_random( lengths2, bits_mask )
# Find intersection
random2 &= bits1
# Print amount intersecting
total_samples[ i, featnum ] += random2.count_range( 0, random2.size )
print >>sys.stderr, total_samples[ i, featnum ]
fraction_overlap = total_samples / total_lengths2
print "\t".join( intervals2_fnames )
print "\t".join( map( str, total_actual/total_lengths2 ) )
for row in fraction_overlap:
print "\t".join( map( str, row ) )
#print "total covered by first: %d, second: %d, overlap: %d" % ( total_lengths1, total_lengths2, total_actual )
print "observed overlap: %d, sample mean: %d, sample stdev: %d" % ( total_actual, stats.amean( total_samples ), stats.asamplestdev( total_samples ) )
print "z-score:", ( total_actual - stats.amean( total_samples ) ) / stats.asamplestdev( total_samples )
print "percentile:", sum( total_actual > total_samples ) / nsamples
def main():
region_fname = sys.argv[1]
mask_fname = sys.argv[2]
nsamples = int( sys.argv[3] )
intervals1_fname = sys.argv[4]
intervals2_fnames = sys.argv[5:]
nfeatures = len( intervals2_fnames )
total_actual = zeros( nfeatures )
# total_lengths1 = 0
total_lengths2 = zeros( nfeatures )
total_samples = zeros( ( nsamples, nfeatures ) )
for line in open( region_fname ):
# Load lengths for all intervals overlapping region
fields = line.split()
print("Processing region:", fields[3], file=sys.stderr)
r_chr, r_start, r_stop = fields[0], int( fields[1] ), int( fields[2] )
r_length = r_stop - r_start
# Load the mask
mask = overlapping_in_bed( mask_fname, r_chr, r_start, r_stop )
bits_mask = as_bits( r_start, r_length, mask )
bits_not_masked = bit_clone( bits_mask ); bits_not_masked.invert()
# Load the first set
intervals1 = overlapping_in_bed( intervals1_fname, r_chr, r_start, r_stop )
bits1 = as_bits( r_start, r_length, intervals1 )
# Intersect it with the mask
bits1.iand( bits_not_masked )
# Sanity checks
assert count_overlap( bits1, bits_mask ) == 0
# For each data set
for featnum, intervals2_fname in enumerate( intervals2_fnames ):
print(intervals2_fname, file=sys.stderr)
intervals2 = overlapping_in_bed( intervals2_fname, r_chr, r_start, r_stop )
bits2 = as_bits( r_start, r_length, intervals2 )
bits2.iand( bits_not_masked )
assert count_overlap( bits2, bits_mask ) == 0
# Observed values
actual_overlap = count_overlap( bits1, bits2 )
total_actual[featnum] += actual_overlap
# Sample
lengths2 = list( interval_lengths( bits2 ) )
total_lengths2[ featnum ] += sum( lengths2 )
for i in range( nsamples ):
# Build randomly covered bitmask for second set
random2 = throw_random( lengths2, bits_mask )
# Find intersection
random2 &= bits1
# Print amount intersecting
total_samples[ i, featnum ] += random2.count_range( 0, random2.size )
print(total_samples[ i, featnum ], file=sys.stderr)
fraction_overlap = total_samples / total_lengths2
print("\t".join( intervals2_fnames ))
print("\t".join( map( str, total_actual/total_lengths2 ) ))
for row in fraction_overlap:
print("\t".join( map( str, row ) ))
#print "total covered by first: %d, second: %d, overlap: %d" % ( total_lengths1, total_lengths2, total_actual )
print("observed overlap: %d, sample mean: %d, sample stdev: %d" % ( total_actual, stats.amean( total_samples ), stats.asamplestdev( total_samples ) ))
print("z-score:", ( total_actual - stats.amean( total_samples ) ) / stats.asamplestdev( total_samples ))
print("percentile:", sum( total_actual > total_samples ) / nsamples)