Mercurial > repos > cstrittmatter > test_eurl_vtec_wgs_pt

annotate scripts/ReMatCh/utils/gffParser.py @ 3:0cbed1c0a762 draft default tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
author cstrittmatter
date Tue, 28 Jan 2020 10:42:31 -0500
parents 965517909457
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
1 #!/usr/bin/env python3
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
2
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
3 import argparse
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
4 import os
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
5 from Bio import SeqIO
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
6 from Bio.Seq import Seq
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
7 from Bio.SeqRecord import SeqRecord
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
8 import ntpath
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
9
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
10 version = '1.0'
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
11
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
12
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
13 def parse_id(filename):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
14 # get wanted feature IDs
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
15 gff_ids = []
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
16 with open(filename, 'r') as in_handle:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
17 for line in in_handle:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
18 line = line.strip()
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
19 gff_ids.append(line)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
20 return gff_ids
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
21
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
22
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
23 def retrieve_seq_file(fasta_file, coord_file, extra_seq, filename, output_dir):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
24 # Parsing the sequence file, using the provided txt file containing the contig ID and positions to retrieve sequences.
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
25 handle = open(fasta_file, "rU")
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
26 records_dict = SeqIO.to_dict(SeqIO.parse(handle, "fasta"))
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
27 handle.close()
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
28
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
29 seq_2_get = {}
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
30 with open(coord_file, 'r') as sequeces2get:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
31 for line in sequeces2get:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
32 line = line.split(',')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
33 coords = (int(line[-2]), int(line[-1]))
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
34 contig_id = line[0]
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
35 if contig_id in list(seq_2_get.keys()):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
36 seq_2_get[contig_id].append(coords)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
37 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
38 seq_2_get[contig_id] = [coords]
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
39
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
40 with open(output_dir + '/' + filename + '.fasta', 'w') as output_handle:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
41 fails = 0
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
42 successes = 0
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
43 records = []
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
44 for contig, listCoords in list(seq_2_get.items()):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
45 contig_seq = records_dict[contig].seq
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
46 for coord in listCoords:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
47 coord1 = coord[0] - extra_seq
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
48 coord2 = coord[1] + extra_seq
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
49 if coord1 < 0 or coord2 > len(contig_seq):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
50 fail_log = open(output_dir + '/' + filename + '_fails.txt', 'a')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
51 fail_log.write(contig + ',' + str(coord[0]) + ',' + str(coord[1]) + '\n')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
52 fail_log.close()
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
53 fails += 1
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
54 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
55 geneseq = str(contig_seq[coord1:coord2])
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
56 record = SeqRecord(Seq(geneseq), id=str(str(contig) + '#' + str(coord1) + '_' + str(coord2)),
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
57 description='')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
58 records.append(record)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
59 successes += 1
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
60 SeqIO.write(records, output_handle, "fasta")
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
61
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
62 print('Retrived %s features successfully from %s with %s bp as extra'
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
63 ' sequence.' % (str(successes), filename, str(extra_seq)))
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
64 if fails > 0:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
65 print('%s featrued failed to retrieve. Check %s_fails.txt file.' % (str(fails), filename))
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
66
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
67
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
68 def retrieve_seq(fasta_file, gff_features, extra_seq, filename, output_dir):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
69 # parsing the sequence file into a SeqIO dictionary. one contig per entry
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
70 handle = open(fasta_file, "rU")
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
71 records_dict = SeqIO.to_dict(SeqIO.parse(handle, "fasta"))
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
72 handle.close()
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
73
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
74 with open(output_dir + '/' + filename + '.fasta', 'w') as output_handle:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
75 fails = 0
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
76 successes = 0
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
77 records = []
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
78 for locus, location in list(gff_features.items()):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
79 # print locus
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
80 contig_seq = records_dict[location[0]].seq
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
81 coord1 = location[1] - extra_seq
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
82 coord2 = location[2] + extra_seq
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
83 if coord1 < 0 or coord2 > len(contig_seq):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
84 fail_log = open(output_dir + '/' + filename + '_fails.txt', 'a')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
85 fail_log.write(locus + '\n')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
86 fail_log.close()
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
87 fails += 1
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
88 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
89 geneseq = str(contig_seq[coord1:coord2])
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
90 if location[3] == '-':
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
91 seq = Seq(geneseq)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
92 geneseq = str(seq.reverse_complement())
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
93 record = SeqRecord(Seq(geneseq),
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
94 id=str(locus + '-' + str(location[0]) + '#' + str(location[1]) + '_' +
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
95 str(location[2])),
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
96 description='')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
97 records.append(record)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
98 successes += 1
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
99 SeqIO.write(records, output_handle, "fasta")
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
100 print('Retrived %s features successfully from %s with %s bp as extra'
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
101 ' sequence.' % (str(successes), filename, str(extra_seq)))
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
102 if fails > 0:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
103 print('%s featrued failed to retrieve. Check %s_fails.txt file.' % (str(fails), filename))
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
104
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
105
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
106 def parse_features(temp_gff):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
107 # parsing the feature file into a dictionary
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
108 gff_features = {}
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
109
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
110 with open(temp_gff, 'r') as temp_genes:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
111 for line in temp_genes:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
112 line = line.split('\t')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
113 if "CDS" in line[2]:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
114 id = line[-1].split(';')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
115 locus_id = str(id[0].split('=')[1])
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
116 contig = line[0]
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
117 begining = int(line[3]) - 1 # to get the full sequence
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
118 end = int(line[4])
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
119 strand = line[6]
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
120 location = [contig, begining, end, strand]
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
121 gff_features[locus_id] = location
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
122 return gff_features
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
123
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
124
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
125 def gff_parser(gff_file, extra_seq=0, output_dir='.', keep_temporary_files=False, ids=None, coord_file=None):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
126 filename = ntpath.basename(gff_file).replace('.gff', '')
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
127
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
128 # cleaning temp files if they exist
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
129 if os.path.isfile(output_dir + '/' + filename + '_features.gff'):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
130 os.remove(output_dir + '/' + filename + '_features.gff')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
131 if os.path.isfile(output_dir + '/' + filename + '_sequence.fasta'):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
132 os.remove(output_dir + '/' + filename + '_sequence.fasta')
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
133
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
134 # cleaning fails file if it exists
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
135 if os.path.isfile(output_dir + '/' + filename + '_fails.txt'):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
136 os.remove(output_dir + '/' + filename + '_fails.txt')
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
137
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
138 if coord_file is None:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
139
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
140 if ids is not None:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
141 select_ids = parse_id(ids)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
142 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
143 select_ids = None
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
144
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
145 # separating the gff into 2 different files: one with the features and another with the conting sequences
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
146 with open(gff_file, 'r') as in_handle, open(output_dir + '/' + filename + '_features.gff', 'a') as temp_genes, \
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
147 open(output_dir + '/' + filename + '_sequence.fasta', 'a') as temp_contigs:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
148 for line in in_handle:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
149 if not line.startswith('##'):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
150 if '\t' in line:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
151 if select_ids is not None:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
152 items = line.split('\t')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
153 id = items[-1].split(';')[0]
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
154 id = id.split('=')[1]
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
155 if id in select_ids:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
156 temp_genes.write(line)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
157 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
158 temp_genes.write(line)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
159 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
160 temp_contigs.write(line)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
161
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
162 gff_files = parse_features(output_dir + '/' + filename + '_features.gff')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
163
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
164 retrieve_seq(output_dir + '/' + filename + '_sequence.fasta', gff_files, extra_seq, filename, output_dir)
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
165
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
166 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
167 with open(gff_file, 'r') as in_handle, \
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
168 open(output_dir + '/' + filename + '_sequence.fasta', 'a') as temp_contigs:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
169 for line in in_handle:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
170 if not line.startswith('##'):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
171 if '\t' in line:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
172 pass
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
173 else:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
174 temp_contigs.write(line)
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
175
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
176 retrieve_seq_file(output_dir + '/' + filename + '_sequence.fasta', coord_file, extra_seq, filename, output_dir)
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
177
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
178 # removing temp files
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
179 if not keep_temporary_files:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
180 try:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
181 os.remove(output_dir + '/' + filename + '_features.gff')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
182 except:
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
183 pass
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
184 os.remove(output_dir + '/' + filename + '_sequence.fasta')
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
185
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
186
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
187 def main():
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
188 parser = argparse.ArgumentParser(prog='gffParser.py', description='GFF3 parser for feature sequence retrival.',
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
189 epilog='by C I Mendes (cimendes@medicina.ulisboa.pt)')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
190 parser.add_argument('--version', help='Version information', action='version', version=str('%(prog)s v' + version))
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
191
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
192 parser.add_argument('-i', '--input',
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
193 help='GFF3 file to parse, containing both sequences and annotations (like the one obtained from'
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
194 ' PROKKA).', type=argparse.FileType('r'), required=True)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
195 parser.add_argument('-x', '--extraSeq', help='Extra sequence to retrieve per feature in gff.', default=0, type=int,
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
196 required=False)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
197 parser.add_argument('-k', '--keepTemporaryFiles', help='Keep temporary gff(without sequence) and fasta files.',
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
198 action='store_true')
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
199 parser.add_argument('-o', '--outputDir', help='Path to where the output is to be saved.', default='.',
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
200 required=False)
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
201
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
202 parser_optional_selected_regions_exclusive = parser.add_mutually_exclusive_group()
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
203 parser_optional_selected_regions_exclusive.add_argument('-s', '--select',
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
204 help='txt file with the IDs of interest, one per line',
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
205 type=argparse.FileType('r'), required=False)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
206 parser_optional_selected_regions_exclusive.add_argument('-f', '--fromFile',
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
207 help='Sequence coordinates to be retrieved. Requires contig'
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
208 ' ID and coords (contig,strart,end) in a csv file. One'
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
209 ' per line.', type=argparse.FileType('r'),
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
210 required=False)
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
211
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
212 args = parser.parse_args()
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
213
3
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
214 args.outputDir = os.path.abspath(args.outputDir)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
215 if not os.path.isdir(args.outputDir):
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
216 os.makedirs(args.outputDir)
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
217
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
218 gff_parser(os.path.abspath(args.input.name), args.extraSeq, os.path.abspath(args.outputDir),
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
219 args.keepTemporaryFiles,
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
220 os.path.abspath(args.select.name) if args.select is not None else None,
0cbed1c0a762 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents: 0
diff changeset
221 os.path.abspath(args.fromFile.name) if args.fromFile is not None else None)
0
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
222
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
223
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
224 if __name__ == "__main__":
965517909457 planemo upload commit 15239f1674081ab51ab8dd75a9a40cf1bfaa93e8
cstrittmatter
parents:
diff changeset
225 main()