annotate find_in_reference.py @ 1:856033fb26e8

Add case insensitive option
author Jim Johnson <jj@umn.edu>
date Fri, 17 Jan 2014 14:50:53 -0600
parents fe0327a3ba81
children 30975b3ff0dc
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
1 #!/usr/bin/env python
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
2 """
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
3 #
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
4 #------------------------------------------------------------------------------
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
5 # University of Minnesota
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
6 # Copyright 2013, Regents of the University of Minnesota
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
7 #------------------------------------------------------------------------------
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
8 # Author:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
9 #
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
10 # James E Johnson
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
11 #
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
12 #------------------------------------------------------------------------------
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
13 """
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
14
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
15 """
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
16 Takes 2 tabular files as input:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
17 1. The file to be filtered
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
18 2. The reference file
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
19
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
20 The string value of selected column of the input file is searched for
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
21 in the string values of the selected column of the reference file.
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
22
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
23 The intended purpose is to filter a peptide fasta file in tabular format
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
24 by whether those peptide sequences are found in a reference fasta file.
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
25
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
26 """
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
27 import sys,re,os.path
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
28 import tempfile
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
29 import optparse
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
30 from optparse import OptionParser
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
31 import logging
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
32
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
33
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
34 def __main__():
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
35 #Parse Command Line
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
36 parser = optparse.OptionParser()
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
37 parser.add_option( '-i', '--input', dest='input', help='The input file to filter. (Otherwise read from stdin)' )
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
38 parser.add_option( '-r', '--reference', dest='reference', help='The reference file to filter against' )
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
39 parser.add_option( '-o', '--output', dest='output', help='The output file for input lines filtered by reference')
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
40 parser.add_option( '-f', '--filtered', dest='filtered', help='The output file for input lines not in the output')
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
41 parser.add_option('-c','--input_column', dest='input_column', default=None, help='The column for the value in the input file. (first column = 1, default to last column)')
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
42 parser.add_option('-C','--reference_column', dest='reference_column', default=None, help='The column for the value in the reference file. (first column = 1, default to last column)')
1
856033fb26e8 Add case insensitive option
Jim Johnson <jj@umn.edu>
parents: 0
diff changeset
43 parser.add_option( '-I', '--case_insensitive', dest='ignore_case', action="store_true", default=False, help='case insensitive' )
0
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
44 parser.add_option( '-k', '--keep', dest='keep', action="store_true", default=False, help='' )
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
45 parser.add_option( '-d', '--debug', dest='debug', action='store_true', default=False, help='Turn on wrapper debugging to stdout' )
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
46 (options, args) = parser.parse_args()
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
47 # Input files
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
48 if options.input != None:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
49 try:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
50 inputPath = os.path.abspath(options.input)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
51 inputFile = open(inputPath, 'r')
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
52 except Exception, e:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
53 print >> sys.stderr, "failed: %s" % e
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
54 exit(2)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
55 else:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
56 inputFile = sys.stdin
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
57 # Reference
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
58 if options.reference == None:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
59 print >> sys.stderr, "failed: reference file is required"
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
60 exit(2)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
61 # Output files
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
62 outFile = None
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
63 filteredFile = None
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
64 if options.filtered == None and options.output == None:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
65 #write to stdout
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
66 outFile = sys.stdout
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
67 else:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
68 if options.output != None:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
69 try:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
70 outPath = os.path.abspath(options.output)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
71 outFile = open(outPath, 'w')
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
72 except Exception, e:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
73 print >> sys.stderr, "failed: %s" % e
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
74 exit(3)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
75 if options.filtered != None:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
76 try:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
77 filteredPath = os.path.abspath(options.filtered)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
78 filteredFile = open(filteredPath, 'w')
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
79 except Exception, e:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
80 print >> sys.stderr, "failed: %s" % e
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
81 exit(3)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
82 incol = -1
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
83 if options.input_column and options.input_column > 0:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
84 incol = int(options.input_column)-1
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
85 refcol = -1
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
86 if options.reference_column and options.reference_column > 0:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
87 refcol = int(options.reference_column)-1
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
88 refFile = None
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
89 for ln,line in enumerate(inputFile):
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
90 try:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
91 found = False
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
92 search_string = line.split('\t')[incol].rstrip('\r\n')
1
856033fb26e8 Add case insensitive option
Jim Johnson <jj@umn.edu>
parents: 0
diff changeset
93 if options.ignore_case:
856033fb26e8 Add case insensitive option
Jim Johnson <jj@umn.edu>
parents: 0
diff changeset
94 search_string = search_string.upper()
0
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
95 if options.debug:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
96 print >> sys.stderr, "search: %s" % (search_string)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
97 refFile = open(options.reference,'r')
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
98 for tn,fline in enumerate(refFile):
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
99 target_string = fline.split('\t')[refcol]
1
856033fb26e8 Add case insensitive option
Jim Johnson <jj@umn.edu>
parents: 0
diff changeset
100 if options.ignore_case:
856033fb26e8 Add case insensitive option
Jim Johnson <jj@umn.edu>
parents: 0
diff changeset
101 target_string = target_string.upper()
0
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
102 if options.debug:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
103 print >> sys.stderr, "in: %s %s %s" % (search_string,search_string in target_string,target_string)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
104 if search_string in target_string:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
105 found = True
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
106 break
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
107 if found:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
108 if options.keep == True:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
109 if outFile:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
110 outFile.write(line)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
111 else:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
112 if filteredFile:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
113 filteredFile.write(line)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
114 else:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
115 if options.keep == True:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
116 if filteredFile:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
117 filteredFile.write(line)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
118 else:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
119 if outFile:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
120 outFile.write(line)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
121 except Exception, e:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
122 print >> sys.stderr, "failed: Error reading %s - %s" % (options.reference,e)
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
123 finally:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
124 if refFile:
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
125 refFile.close()
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
126
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
127 if __name__ == "__main__" : __main__()
fe0327a3ba81 Uploaded
jjohnson
parents:
diff changeset
128