0
|
1 #!/usr/bin/env python
|
|
2 #Guruprasad Ananda
|
|
3 """
|
|
4 Calculate count and coverage of one query on another, and append the Coverage and counts to
|
|
5 the last four columns as bases covered, percent coverage, number of completely present features, number of partially present/overlapping features.
|
|
6
|
|
7 usage: %prog bed_file_1 bed_file_2 out_file
|
|
8 -1, --cols1=N,N,N,N: Columns for chr, start, end, strand in first file
|
|
9 -2, --cols2=N,N,N,N: Columns for chr, start, end, strand in second file
|
|
10 """
|
|
11 import sys, fileinput
|
|
12 from bx.intervals.io import *
|
|
13 from bx.cookbook import doc_optparse
|
|
14 from bx.intervals.operations import quicksect
|
|
15 from galaxy.tools.util.galaxyops import *
|
|
16
|
|
17 assert sys.version_info[:2] >= ( 2, 4 )
|
|
18
|
|
19 def stop_err(msg):
|
|
20 sys.stderr.write(msg)
|
|
21 sys.exit()
|
|
22
|
|
23 def counter(node, start, end):
|
|
24 global full, partial
|
|
25 if node.start <= start and node.maxend > start:
|
|
26 if node.end >= end or (node.start == start and end > node.end > start):
|
|
27 full += 1
|
|
28 elif end > node.end > start:
|
|
29 partial += 1
|
|
30 if node.left and node.left.maxend > start:
|
|
31 counter(node.left, start, end)
|
|
32 if node.right:
|
|
33 counter(node.right, start, end)
|
|
34 elif start < node.start < end:
|
|
35 if node.end <= end:
|
|
36 full += 1
|
|
37 else:
|
|
38 partial += 1
|
|
39 if node.left and node.left.maxend > start:
|
|
40 counter(node.left, start, end)
|
|
41 if node.right:
|
|
42 counter(node.right, start, end)
|
|
43 else:
|
|
44 if node.left:
|
|
45 counter(node.left, start, end)
|
|
46
|
|
47 def count_coverage( readers, comments=True ):
|
|
48 primary = readers[0]
|
|
49 secondary = readers[1]
|
|
50 secondary_copy = readers[2]
|
|
51
|
|
52 rightTree = quicksect.IntervalTree()
|
|
53 for item in secondary:
|
|
54 if type( item ) is GenomicInterval:
|
|
55 rightTree.insert( item, secondary.linenum, item.fields )
|
|
56
|
|
57 bitsets = secondary_copy.binned_bitsets()
|
|
58
|
|
59 global full, partial
|
|
60
|
|
61 for interval in primary:
|
|
62 if type( interval ) is Header:
|
|
63 yield interval
|
|
64 if type( interval ) is Comment and comments:
|
|
65 yield interval
|
|
66 elif type( interval ) == GenomicInterval:
|
|
67 chrom = interval.chrom
|
|
68 start = int(interval.start)
|
|
69 end = int(interval.end)
|
|
70 full = 0
|
|
71 partial = 0
|
|
72 if chrom not in bitsets:
|
|
73 bases_covered = 0
|
|
74 percent = 0.0
|
|
75 full = 0
|
|
76 partial = 0
|
|
77 else:
|
|
78 bases_covered = bitsets[ chrom ].count_range( start, end-start )
|
|
79 if (end - start) == 0:
|
|
80 percent = 0
|
|
81 else:
|
|
82 percent = float(bases_covered) / float(end - start)
|
|
83 if bases_covered:
|
|
84 root = rightTree.chroms[chrom] #root node for the chrom tree
|
|
85 counter(root, start, end)
|
|
86 interval.fields.append(str(bases_covered))
|
|
87 interval.fields.append(str(percent))
|
|
88 interval.fields.append(str(full))
|
|
89 interval.fields.append(str(partial))
|
|
90 yield interval
|
|
91
|
|
92
|
|
93 def main():
|
|
94 options, args = doc_optparse.parse( __doc__ )
|
|
95
|
|
96 try:
|
|
97 chr_col_1, start_col_1, end_col_1, strand_col_1 = parse_cols_arg( options.cols1 )
|
|
98 chr_col_2, start_col_2, end_col_2, strand_col_2 = parse_cols_arg( options.cols2 )
|
|
99 in1_fname, in2_fname, out_fname = args
|
|
100 except:
|
|
101 stop_err( "Data issue: click the pencil icon in the history item to correct the metadata attributes." )
|
|
102
|
|
103 g1 = NiceReaderWrapper( fileinput.FileInput( in1_fname ),
|
|
104 chrom_col=chr_col_1,
|
|
105 start_col=start_col_1,
|
|
106 end_col=end_col_1,
|
|
107 strand_col=strand_col_1,
|
|
108 fix_strand=True )
|
|
109 g2 = NiceReaderWrapper( fileinput.FileInput( in2_fname ),
|
|
110 chrom_col=chr_col_2,
|
|
111 start_col=start_col_2,
|
|
112 end_col=end_col_2,
|
|
113 strand_col=strand_col_2,
|
|
114 fix_strand=True )
|
|
115 g2_copy = NiceReaderWrapper( fileinput.FileInput( in2_fname ),
|
|
116 chrom_col=chr_col_2,
|
|
117 start_col=start_col_2,
|
|
118 end_col=end_col_2,
|
|
119 strand_col=strand_col_2,
|
|
120 fix_strand=True )
|
|
121
|
|
122
|
|
123 out_file = open( out_fname, "w" )
|
|
124
|
|
125 try:
|
|
126 for line in count_coverage([g1, g2, g2_copy]):
|
|
127 if type( line ) is GenomicInterval:
|
|
128 out_file.write( "%s\n" % "\t".join( line.fields ) )
|
|
129 else:
|
|
130 out_file.write( "%s\n" % line )
|
|
131 except ParseError, exc:
|
|
132 out_file.close()
|
|
133 fail( str( exc ) )
|
|
134
|
|
135 out_file.close()
|
|
136
|
|
137 if g1.skipped > 0:
|
|
138 print skipped( g1, filedesc=" of 1st dataset" )
|
|
139 if g2.skipped > 0:
|
|
140 print skipped( g2, filedesc=" of 2nd dataset" )
|
|
141 elif g2_copy.skipped > 0:
|
|
142 print skipped( g2_copy, filedesc=" of 2nd dataset" )
|
|
143
|
|
144
|
|
145 if __name__ == "__main__":
|
|
146 main()
|