Mercurial > repos > earlhaminst > gstf_preparation

annotate gstf_preparation.py @ 3:7e11a7f4bdba draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 77ed525b753e34d3e9dd4f07a239592ce764f7e6-dirty
author earlhaminst
date Fri, 24 Nov 2017 12:32:39 -0500
parents a36645976342
children 284f64ad9d43
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
1 from __future__ import print_function
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
2
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
3 import collections
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
4 import json
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
5 import optparse
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
6 import sqlite3
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
7 import sys
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
8
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
9 version = "0.3.0"
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
10 gene_count = 0
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
11
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
12 Sequence = collections.namedtuple('Sequence', ['header', 'sequence'])
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
13
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
14
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
15 def FASTAReader_gen(fasta_filename):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
16 with open(fasta_filename) as fasta_file:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
17 line = fasta_file.readline()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
18 while True:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
19 if not line:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
20 return
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
21 assert line.startswith('>'), "FASTA headers must start with >"
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
22 header = line.rstrip()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
23 sequence_parts = []
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
24 line = fasta_file.readline()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
25 while line and line[0] != '>':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
26 sequence_parts.append(line.rstrip())
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
27 line = fasta_file.readline()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
28 sequence = "\n".join(sequence_parts)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
29 yield Sequence(header, sequence)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
30
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
31
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
32 def create_tables(conn):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
33 cur = conn.cursor()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
34
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
35 cur.execute('''CREATE TABLE meta (
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
36 version VARCHAR PRIMARY KEY NOT NULL)''')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
37
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
38 cur.execute('INSERT INTO meta (version) VALUES (?)',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
39 (version, ))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
40
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
41 cur.execute('''CREATE TABLE gene (
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
42 gene_id VARCHAR PRIMARY KEY NOT NULL,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
43 gene_symbol VARCHAR,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
44 species VARCHAR NOT NULL,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
45 gene_json VARCHAR NOT NULL)''')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
46 cur.execute('CREATE INDEX gene_symbol_index ON gene (gene_symbol)')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
47
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
48 cur.execute('''CREATE TABLE transcript (
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
49 transcript_id VARCHAR PRIMARY KEY NOT NULL,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
50 protein_id VARCHAR UNIQUE,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
51 protein_sequence VARCHAR,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
52 gene_id VARCHAR NOT NULL REFERENCES gene(gene_id))''')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
53
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
54 cur.execute('''CREATE VIEW transcript_species AS
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
55 SELECT transcript_id, species
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
56 FROM transcript JOIN gene
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
57 ON transcript.gene_id = gene.gene_id''')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
58
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
59 conn.commit()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
60
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
61
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
62 def remove_type_from_list_of_ids(l):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
63 return ','.join(remove_type_from_id(_) for _ in l.split(','))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
64
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
65
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
66 def remove_type_from_id(id_):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
67 colon_index = id_.find(':')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
68 if colon_index >= 0:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
69 return id_[colon_index + 1:]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
70 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
71 return id_
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
72
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
73
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
74 def feature_to_dict(cols, parent_dict=None):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
75 d = {
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
76 'end': int(cols[4]),
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
77 'start': int(cols[3]),
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
78 }
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
79 for attr in cols[8].split(';'):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
80 if '=' in attr:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
81 (tag, value) = attr.split('=')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
82 if tag == 'ID':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
83 tag = 'id'
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
84 value = remove_type_from_id(value)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
85 elif tag == 'Parent':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
86 value = remove_type_from_list_of_ids(value)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
87 d[tag] = value
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
88 if cols[6] == '+':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
89 d['strand'] = 1
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
90 elif cols[6] == '-':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
91 d['strand'] = -1
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
92 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
93 raise Exception("Unrecognized strand '%s'" % cols[6])
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
94 if parent_dict is not None and 'Parent' in d:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
95 # a 3' UTR can be split among multiple exons
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
96 # a 5' UTR can be split among multiple exons
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
97 # a CDS can be part of multiple transcripts
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
98 for parent in d['Parent'].split(','):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
99 if parent not in parent_dict:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
100 parent_dict[parent] = [d]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
101 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
102 parent_dict[parent].append(d)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
103 return d
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
104
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
105
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
106 def add_gene_to_dict(cols, species, gene_dict):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
107 global gene_count
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
108 gene = feature_to_dict(cols)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
109 gene.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
110 'member_id': gene_count,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
111 'object_type': 'Gene',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
112 'seq_region_name': cols[0],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
113 'species': species,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
114 'Transcript': [],
1
a36645976342 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 68cf06ec77e20eaf057d77d5bea19e201c4ab27e-dirty
earlhaminst
parents: 0
diff changeset
115 'display_name': gene.get('Name', None)
0
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
116 })
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
117 if gene['id']:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
118 gene_dict[gene['id']] = gene
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
119 gene_count = gene_count + 1
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
120
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
121
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
122 def add_transcript_to_dict(cols, species, transcript_dict):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
123 transcript = feature_to_dict(cols)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
124 transcript.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
125 'object_type': 'Transcript',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
126 'seq_region_name': cols[0],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
127 'species': species,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
128 })
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
129 transcript_dict[transcript['id']] = transcript
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
130
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
131
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
132 def add_exon_to_dict(cols, species, exon_parent_dict):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
133 exon = feature_to_dict(cols, exon_parent_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
134 exon.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
135 'length': int(cols[4]) - int(cols[3]) + 1,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
136 'object_type': 'Exon',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
137 'seq_region_name': cols[0],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
138 'species': species,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
139 })
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
140 if 'id' not in exon and 'Name' in exon:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
141 exon['id'] = exon['Name']
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
142
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
143
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
144 def add_cds_to_dict(cols, cds_parent_dict):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
145 cds = feature_to_dict(cols, cds_parent_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
146 if 'id' not in cds:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
147 if 'Name' in cds:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
148 cds['id'] = cds['Name']
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
149 elif 'Parent' in cds and ',' not in cds['Parent']:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
150 cds['id'] = cds['Parent']
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
151
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
152
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
153 def join_dicts(gene_dict, transcript_dict, exon_parent_dict, cds_parent_dict, five_prime_utr_parent_dict, three_prime_utr_parent_dict):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
154 for parent, exon_list in exon_parent_dict.items():
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
155 if parent in transcript_dict:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
156 exon_list.sort(key=lambda _: _['start'])
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
157 transcript_dict[parent]['Exon'] = exon_list
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
158
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
159 for transcript_id, transcript in transcript_dict.items():
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
160 translation = {
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
161 'CDS': [],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
162 'id': None,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
163 'end': transcript['end'],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
164 'object_type': 'Translation',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
165 'species': transcript['species'],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
166 'start': transcript['start'],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
167 }
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
168 found_cds = False
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
169 derived_translation_start = None
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
170 derived_translation_end = None
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
171 if transcript_id in cds_parent_dict:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
172 cds_list = cds_parent_dict[transcript_id]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
173 cds_ids = set(_['id'] for _ in cds_list)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
174 if len(cds_ids) > 1:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
175 raise Exception("Transcript %s has multiple CDSs: this is not supported by Ensembl JSON format" % parent)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
176 translation['id'] = cds_ids.pop()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
177 cds_list.sort(key=lambda _: _['start'])
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
178 translation['CDS'] = cds_list
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
179 translation['start'] = cds_list[0]['start']
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
180 translation['end'] = cds_list[-1]['end']
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
181 found_cds = True
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
182 if transcript_id in five_prime_utr_parent_dict:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
183 five_prime_utr_list = five_prime_utr_parent_dict[transcript_id]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
184 five_prime_utr_list.sort(key=lambda _: _['start'])
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
185 if transcript['strand'] == 1:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
186 derived_translation_start = five_prime_utr_list[-1]['end'] + 1
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
187 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
188 derived_translation_end = five_prime_utr_list[0]['start'] - 1
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
189 if transcript_id in three_prime_utr_parent_dict:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
190 three_prime_utr_list = three_prime_utr_parent_dict[transcript_id]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
191 three_prime_utr_list.sort(key=lambda _: _['start'])
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
192 if transcript['strand'] == 1:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
193 derived_translation_end = three_prime_utr_list[0]['start'] - 1
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
194 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
195 derived_translation_start = three_prime_utr_list[-1]['end'] + 1
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
196 if derived_translation_start is not None:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
197 if found_cds:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
198 if derived_translation_start > translation['start']:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
199 raise Exception("UTR overlaps with CDS")
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
200 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
201 translation['start'] = derived_translation_start
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
202 if derived_translation_end is not None:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
203 if found_cds:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
204 if derived_translation_end < translation['end']:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
205 raise Exception("UTR overlaps with CDS")
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
206 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
207 translation['end'] = derived_translation_end
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
208 if found_cds or derived_translation_start is not None or derived_translation_end is not None:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
209 transcript['Translation'] = translation
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
210
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
211 for transcript in transcript_dict.values():
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
212 if 'Parent' in transcript:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
213 # A polycistronic transcript can have multiple parents
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
214 for parent in transcript['Parent'].split(','):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
215 if parent in gene_dict:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
216 gene_dict[parent]['Transcript'].append(transcript)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
217
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
218
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
219 def write_gene_dict_to_db(conn, gene_dict):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
220 cur = conn.cursor()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
221
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
222 for gene in gene_dict.values():
3
7e11a7f4bdba planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 77ed525b753e34d3e9dd4f07a239592ce764f7e6-dirty
earlhaminst
parents: 1
diff changeset
223 if gene is None:
7e11a7f4bdba planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 77ed525b753e34d3e9dd4f07a239592ce764f7e6-dirty
earlhaminst
parents: 1
diff changeset
224 # This can happen when loading a JSON file from Ensembl
7e11a7f4bdba planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 77ed525b753e34d3e9dd4f07a239592ce764f7e6-dirty
earlhaminst
parents: 1
diff changeset
225 continue
0
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
226 gene_id = gene['id']
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
227 cur.execute('INSERT INTO gene (gene_id, gene_symbol, species, gene_json) VALUES (?, ?, ?, ?)',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
228 (gene_id, gene.get('display_name', None), gene['species'], json.dumps(gene)))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
229
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
230 if "Transcript" in gene:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
231 for transcript in gene["Transcript"]:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
232 transcript_id = transcript['id']
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
233 protein_id = transcript.get('Translation', {}).get('id', None)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
234 try:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
235 cur.execute('INSERT INTO transcript (transcript_id, protein_id, gene_id) VALUES (?, ?, ?)',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
236 (transcript_id, protein_id, gene_id))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
237 except Exception as e:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
238 raise Exception("Error while inserting (%s, %s, %s) into transcript table: %s" % (transcript_id, protein_id, gene_id, e))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
239
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
240 conn.commit()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
241
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
242
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
243 def fetch_species_for_transcript(conn, transcript_id):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
244 cur = conn.cursor()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
245
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
246 cur.execute('SELECT species FROM transcript_species WHERE transcript_id=?',
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
247 (transcript_id, ))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
248 results = cur.fetchone()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
249 if not results:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
250 return None
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
251 return results[0]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
252
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
253
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
254 def remove_id_version(s):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
255 """
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
256 Remove the optional '.VERSION' from an Ensembl id.
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
257 """
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
258 if s.startswith('ENS'):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
259 return s.split('.')[0]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
260 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
261 return s
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
262
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
263
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
264 def __main__():
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
265 parser = optparse.OptionParser()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
266 parser.add_option('--gff3', action='append', default=[], help='GFF3 file to convert, in SPECIES:FILENAME format. Use multiple times to add more files')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
267 parser.add_option('--json', action='append', default=[], help='JSON file to merge. Use multiple times to add more files')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
268 parser.add_option('--fasta', action='append', default=[], help='Path of the input FASTA files')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
269 parser.add_option('-o', '--output', help='Path of the output SQLite file')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
270 parser.add_option('--of', help='Path of the output FASTA file')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
271 options, args = parser.parse_args()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
272 if args:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
273 raise Exception('Use options to provide inputs')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
274
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
275 conn = sqlite3.connect(options.output)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
276 conn.execute('PRAGMA foreign_keys = ON')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
277 create_tables(conn)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
278
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
279 for gff3_arg in options.gff3:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
280 try:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
281 (species, filename) = gff3_arg.split(':')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
282 except ValueError:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
283 raise Exception("Argument for --gff3 '%s' is not in the SPECIES:FILENAME format" % gff3_arg)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
284 gene_dict = dict()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
285 transcript_dict = dict()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
286 exon_parent_dict = dict()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
287 cds_parent_dict = dict()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
288 five_prime_utr_parent_dict = dict()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
289 three_prime_utr_parent_dict = dict()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
290 with open(filename) as f:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
291 for i, line in enumerate(f, start=1):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
292 line = line.strip()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
293 if not line:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
294 # skip empty lines
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
295 continue
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
296 if line[0] == '#':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
297 # skip comment lines
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
298 continue
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
299 cols = line.split('\t')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
300 if len(cols) != 9:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
301 raise Exception("Line %i in file '%s': '%s' does not have 9 columns" % (i, filename, line))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
302 feature_type = cols[2]
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
303 try:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
304 if feature_type == 'gene':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
305 add_gene_to_dict(cols, species, gene_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
306 elif feature_type in ('mRNA', 'transcript'):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
307 add_transcript_to_dict(cols, species, transcript_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
308 elif feature_type == 'exon':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
309 add_exon_to_dict(cols, species, exon_parent_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
310 elif feature_type == 'five_prime_UTR':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
311 feature_to_dict(cols, five_prime_utr_parent_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
312 elif feature_type == 'three_prime_UTR':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
313 feature_to_dict(cols, three_prime_utr_parent_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
314 elif feature_type == 'CDS':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
315 add_cds_to_dict(cols, cds_parent_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
316 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
317 print("Line %i in file '%s': '%s' is not an implemented feature type" % (i, filename, feature_type), file=sys.stderr)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
318 except Exception as e:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
319 print("Line %i in file '%s': %s" % (i, filename, e), file=sys.stderr)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
320
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
321 join_dicts(gene_dict, transcript_dict, exon_parent_dict, cds_parent_dict, five_prime_utr_parent_dict, three_prime_utr_parent_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
322 write_gene_dict_to_db(conn, gene_dict)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
323
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
324 for json_arg in options.json:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
325 with open(json_arg) as f:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
326 write_gene_dict_to_db(conn, json.load(f))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
327
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
328 with open(options.of, 'w') as output_fasta_file:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
329 for fasta_arg in options.fasta:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
330 for entry in FASTAReader_gen(fasta_arg):
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
331 # Extract the transcript id by removing everything after the first space and then removing the version if it is an Ensembl id
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
332 transcript_id = remove_id_version(entry.header[1:].lstrip().split(' ')[0])
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
333 species_for_transcript = fetch_species_for_transcript(conn, transcript_id)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
334 if not species_for_transcript:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
335 print("Transcript '%s' not found in the gene feature information" % transcript_id, file=sys.stderr)
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
336 continue
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
337 # Remove any underscore in the species
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
338 species_for_transcript = species_for_transcript.replace('_', '')
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
339 # Write the FASTA sequence using '>TranscriptId_species' as the header, as required by TreeBest
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
340 output_fasta_file.write(">%s_%s\n%s\n" % (transcript_id, species_for_transcript, entry.sequence))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
341
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
342 conn.close()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
343
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
344
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
345 if __name__ == '__main__':
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670
earlhaminst
parents:
diff changeset
346 __main__()