Mercurial > repos > peterjc > blastxml_to_top_descr

comparison blastxml_to_top_descr/blastxml_to_top_descr.py @ 9:6aafa0ced802 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Uploaded v0.0.8, moved development to GitHub; no functional changes
author peterjc
date Tue, 30 Jul 2013 08:07:28 -0400
parents
children
comparison
equal deleted inserted replaced
8:a8ef75aab1f9 9:6aafa0ced802
1 #!/usr/bin/env python
2 """Convert a BLAST XML file to a top hits description table.
3
4 Takes three command line options, input BLAST XML filename, output tabular
5 BLAST filename, number of hits to collect the descriptions of.
6 """
7 import sys
8 import re
9
10 if "-v" in sys.argv or "--version" in sys.argv:
11 print "v0.0.5"
12 sys.exit(0)
13
14 if sys.version_info[:2] >= ( 2, 5 ):
15 import xml.etree.cElementTree as ElementTree
16 else:
17 from galaxy import eggs
18 import pkg_resources; pkg_resources.require( "elementtree" )
19 from elementtree import ElementTree
20
21 def stop_err( msg ):
22 sys.stderr.write("%s\n" % msg)
23 sys.exit(1)
24
25 #Parse Command Line
26 try:
27 in_file, out_file, topN = sys.argv[1:]
28 except:
29 stop_err("Expect 3 arguments: input BLAST XML file, output tabular file, number of hits")
30
31
32 try:
33 topN = int(topN)
34 except ValueError:
35 stop_err("Number of hits argument should be an integer (at least 1)")
36 if topN < 1:
37 stop_err("Number of hits argument should be an integer (at least 1)")
38
39 # get an iterable
40 try:
41 context = ElementTree.iterparse(in_file, events=("start", "end"))
42 except:
43 stop_err("Invalid data format.")
44 # turn it into an iterator
45 context = iter(context)
46 # get the root element
47 try:
48 event, root = context.next()
49 except:
50 stop_err( "Invalid data format." )
51
52
53 re_default_query_id = re.compile("^Query_\d+$")
54 assert re_default_query_id.match("Query_101")
55 assert not re_default_query_id.match("Query_101a")
56 assert not re_default_query_id.match("MyQuery_101")
57 re_default_subject_id = re.compile("^Subject_\d+$")
58 assert re_default_subject_id.match("Subject_1")
59 assert not re_default_subject_id.match("Subject_")
60 assert not re_default_subject_id.match("Subject_12a")
61 assert not re_default_subject_id.match("TheSubject_1")
62
63
64 count = 0
65 pos_count = 0
66 outfile = open(out_file, 'w')
67 outfile.write("#Query\t%s\n" % "\t".join("BLAST hit %i" % (i+1) for i in range(topN)))
68 for event, elem in context:
69 # for every <Iteration> tag
70 if event == "end" and elem.tag == "Iteration":
71 #Expecting either this, from BLAST 2.2.25+ using FASTA vs FASTA
72 # <Iteration_query-ID>sp|Q9BS26|ERP44_HUMAN</Iteration_query-ID>
73 # <Iteration_query-def>Endoplasmic reticulum resident protein 44 OS=Homo sapiens GN=ERP44 PE=1 SV=1</Iteration_query-def>
74 # <Iteration_query-len>406</Iteration_query-len>
75 # <Iteration_hits></Iteration_hits>
76 #
77 #Or, from BLAST 2.2.24+ run online
78 # <Iteration_query-ID>Query_1</Iteration_query-ID>
79 # <Iteration_query-def>Sample</Iteration_query-def>
80 # <Iteration_query-len>516</Iteration_query-len>
81 # <Iteration_hits>...
82 qseqid = elem.findtext("Iteration_query-ID")
83 if qseqid is None:
84 stop_err("Missing <Iteration_query-ID> (could be really old BLAST XML data?)")
85 if re_default_query_id.match(qseqid):
86 #Place holder ID, take the first word of the query definition
87 qseqid = elem.findtext("Iteration_query-def").split(None,1)[0]
88 # for every <Hit> within <Iteration>
89 hit_descrs = []
90 for hit in elem.findall("Iteration_hits/Hit"):
91 #Expecting either this,
92 # <Hit_id>gi|3024260|sp|P56514.1|OPSD_BUFBU</Hit_id>
93 # <Hit_def>RecName: Full=Rhodopsin</Hit_def>
94 # <Hit_accession>P56514</Hit_accession>
95 #or,
96 # <Hit_id>Subject_1</Hit_id>
97 # <Hit_def>gi|57163783|ref|NP_001009242.1| rhodopsin [Felis catus]</Hit_def>
98 # <Hit_accession>Subject_1</Hit_accession>
99 #
100 #apparently depending on the parse_deflines switch
101 sseqid = hit.findtext("Hit_id").split(None,1)[0]
102 hit_def = sseqid + " " + hit.findtext("Hit_def")
103 if re_default_subject_id.match(sseqid) \
104 and sseqid == hit.findtext("Hit_accession"):
105 #Place holder ID, take the first word of the subject definition
106 hit_def = hit.findtext("Hit_def")
107 sseqid = hit_def.split(None,1)[0]
108 assert hit_def not in hit_descrs
109 hit_descrs.append(hit_def)
110 #print "%r has %i hits" % (qseqid, len(hit_descrs))
111 if hit_descrs:
112 pos_count += 1
113 hit_descrs = hit_descrs[:topN]
114 while len(hit_descrs) < topN:
115 hit_descrs.append("")
116 outfile.write("%s\t%s\n" % (qseqid, "\t".join(hit_descrs)))
117 count += 1
118 # prevents ElementTree from growing large datastructure
119 root.clear()
120 elem.clear()
121 outfile.close()
122 print "Of %i queries, %i had BLAST results" % (count, pos_count)