gstf_preparation: gstf_preparation.py annotate

annotate gstf_preparation.py @ 11:dbe37a658cd2 draft

"planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812"

author	earlhaminst
date	Sun, 27 Sep 2020 18:54:31 +0000
parents	e8e75a79de59
children	99bae410128c

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 json
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	4 import optparse
10 e8e75a79de59 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9c8611fee927883f50bc6955771aa69df1ce8457" earlhaminst parents: 9 diff changeset	5 import os
0 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
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	9 version = "0.5.0"
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
8 92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	12
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	13 class Sequence(object):
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	14 def __init__(self, header, sequence_parts):
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	15 self.header = header
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	16 self.sequence_parts = sequence_parts
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	17 self._sequence = None
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	18
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	19 @property
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	20 def sequence(self):
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	21 if self._sequence is None:
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	22 self._sequence = ''.join(self.sequence_parts)
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	23 return self._sequence
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	24
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	25 def print(self, fh=sys.stdout):
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	26 print(self.header, file=fh)
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	27 for line in self.sequence_parts:
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	28 print(line, file=fh)
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	29
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 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	32 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	33 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	34 while True:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	35 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	36 return
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	37 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	38 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	39 sequence_parts = []
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	40 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	41 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	42 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	43 line = fasta_file.readline()
8 92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	44 yield Sequence(header, sequence_parts)
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	45
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	46
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	47 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	48 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	49
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	50 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	51 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	52
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	53 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	54 (version, ))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	55
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	56 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	57 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	58 gene_symbol VARCHAR,
6 56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	59 seq_region_name VARCHAR NOT NULL,
56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	60 seq_region_start INTEGER NOT NULL,
56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	61 seq_region_end INTEGER NOT NULL,
56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	62 seq_region_strand INTEGER NOT NULL,
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	63 species VARCHAR NOT NULL,
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	64 biotype VARCHAR,
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	65 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	66 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	67
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	68 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	69 transcript_id VARCHAR PRIMARY KEY NOT NULL,
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	70 transcript_symbol VARCHAR,
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	71 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	72 protein_sequence VARCHAR,
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	73 biotype VARCHAR,
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	74 is_canonical BOOLEAN NOT NULL DEFAULT FALSE,
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	75 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	76
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	77 # The following temporary view is not used in GAFA, so schema changes to it
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	78 # don't require a meta version upgrade.
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	79 cur.execute('''CREATE TEMPORARY VIEW transcript_join_gene AS
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	80 SELECT transcript_id, transcript_symbol, COALESCE(transcript.biotype, gene.biotype) AS biotype, is_canonical, gene_id, gene_symbol, seq_region_name, species
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	81 FROM transcript JOIN gene
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	82 USING (gene_id)''')
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	83
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	84 conn.commit()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	85
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	86
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	87 def fetch_transcript_and_gene(conn, transcript_id):
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	88 cur = conn.cursor()
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	89
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	90 cur.execute('SELECT * FROM transcript_join_gene WHERE transcript_id=?',
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	91 (transcript_id, ))
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	92 return cur.fetchone()
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	93
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	94
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	95 def remove_type_from_list_of_ids(ids):
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	96 return ','.join(remove_type_from_id(id_) for id_ in ids.split(','))
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	97
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	98
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	99 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	100 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	101 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	102 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	103 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	104 return id_
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
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	107 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	108 d = {
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	109 '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	110 '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	111 }
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	112 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	113 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	114 (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	115 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	116 tag = 'id'
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	117 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	118 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	119 value = remove_type_from_list_of_ids(value)
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	120 elif tag == 'representative':
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	121 tag = 'is_canonical'
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	122 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	123 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	124 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	125 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	126 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	127 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	128 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	129 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	130 # 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	131 # 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	132 # 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	133 for parent in d['Parent'].split(','):
10 e8e75a79de59 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9c8611fee927883f50bc6955771aa69df1ce8457" earlhaminst parents: 9 diff changeset	134 parent_dict.setdefault(parent, []).append(d)
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	135 return d
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	136
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	137
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	138 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	139 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	140 gene = feature_to_dict(cols)
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	141 if not gene['id']:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	142 raise Exception("Id not found among column 9 attribute tags: %s" % cols[8])
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	143 gene.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	144 '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	145 '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	146 '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	147 'species': species,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	148 'Transcript': [],
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	149 'display_name': gene.get('Name'),
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	150 })
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	151 gene_dict[gene['id']] = gene
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	152 gene_count = gene_count + 1
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	153
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	154
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	155 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	156 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	157 transcript.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	158 '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	159 '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	160 'species': species,
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	161 'display_name': transcript.get('Name'),
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	162 })
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	163 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	164
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	165
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	166 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	167 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	168 exon.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	169 '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	170 '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	171 '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	172 'species': species,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	173 })
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	174 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	175 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	176
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	177
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	178 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	179 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	180 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	181 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	182 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	183 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	184 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	185
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	186
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	187 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	188 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	189 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	190 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	191 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	192
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	193 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	194 translation = {
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	195 'CDS': [],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	196 'id': None,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	197 '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	198 '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	199 '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	200 '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	201 }
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	202 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	203 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	204 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	205 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	206 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	207 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	208 if len(cds_ids) > 1:
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	209 raise Exception("Transcript %s has multiple CDSs: this is not supported by Ensembl JSON format" % transcript_id)
b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	210 cds_id = cds_ids.pop()
b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	211 translation['id'] = cds_id
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	212 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	213 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	214 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	215 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	216 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	217 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	218 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	219 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	220 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	221 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	222 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	223 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	224 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	225 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	226 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	227 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	228 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	229 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	230 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	231 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	232 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	233 if derived_translation_start > translation['start']:
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	234 raise Exception("Transcript %s has the start of CDS %s overlapping with the UTR end" % (transcript_id, cds_id))
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	235 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	236 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	237 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	238 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	239 if derived_translation_end < translation['end']:
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	240 raise Exception("Transcript %s has the end of CDS %s overlapping with the UTR start" % (transcript_id, cds_id))
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	241 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	242 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	243 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	244 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	245
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	246 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	247 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	248 # 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	249 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	250 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	251 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	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 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	255 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	256
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	257 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	258 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	259 # 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	260 continue
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	261 if 'confidence' in gene and gene['confidence'] != 'high':
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	262 print("Gene %s has confidence %s (not high), discarding" % (gene['id'], gene['confidence']), file=sys.stderr)
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	263 continue
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	264 gene_id = gene['id']
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	265 cur.execute('INSERT INTO gene (gene_id, gene_symbol, seq_region_name, seq_region_start, seq_region_end, seq_region_strand, species, biotype, gene_json) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)',
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	266 (gene_id, gene.get('display_name'), gene['seq_region_name'], gene['start'], gene['end'], gene['strand'], gene['species'], gene.get('biotype'), json.dumps(gene)))
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	267
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	268 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	269 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	270 transcript_id = transcript['id']
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	271 transcript_symbol = transcript.get('display_name')
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	272 protein_id = transcript.get('Translation', {}).get('id')
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	273 biotype = transcript.get('biotype')
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	274 is_canonical = transcript.get('is_canonical', False)
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	275 to_insert = (transcript_id, transcript_symbol, protein_id, biotype, is_canonical, gene_id)
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	276 try:
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	277 cur.execute('INSERT INTO transcript (transcript_id, transcript_symbol, protein_id, biotype, is_canonical, gene_id) VALUES (?, ?, ?, ?, ?, ?)',
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	278 to_insert)
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	279 except Exception as e:
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	280 raise Exception("Error while inserting %s into transcript table: %s" % (str(to_insert), e))
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	281
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	282 conn.commit()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	283
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	284
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	285 def remove_id_version(s, force=False):
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	286 """
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	287 Remove the optional '.VERSION' from an id if it's an Ensembl id or if
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	288 `force` is True.
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	289 """
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	290 if force or s.startswith('ENS'):
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	291 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	292 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	293 return s
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	294
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	295
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	296 def __main__():
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	297 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	298 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	299 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	300 parser.add_option('--fasta', action='append', default=[], help='Path of the input FASTA files')
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	301 parser.add_option('--filter', type='choice', choices=['canonical', 'coding', ''], default='', help='Which transcripts to keep')
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	302 parser.add_option('--headers', type='choice',
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	303 choices=['TranscriptId_species', 'GeneSymbol-TranscriptID_species', 'TranscriptSymbol-TranscriptID_species', ''],
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	304 default='', help='Change the header line of the FASTA sequences to this format')
6 56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	305 parser.add_option('--regions', default="", help='Comma-separated list of region IDs for which FASTA sequences should be filtered')
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	306 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	307 parser.add_option('--of', help='Path of the output FASTA file')
10 e8e75a79de59 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9c8611fee927883f50bc6955771aa69df1ce8457" earlhaminst parents: 9 diff changeset	308 parser.add_option('--ff', default=os.devnull, help='Path of the filtered sequences output FASTA file')
6 56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	309
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	310 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	311 if args:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	312 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	313
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	314 conn = sqlite3.connect(options.output)
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	315 conn.row_factory = sqlite3.Row
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	316 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	317 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	318
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	319 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	320 try:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	321 (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	322 except ValueError:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	323 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	324 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	325 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	326 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	327 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	328 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	329 three_prime_utr_parent_dict = dict()
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	330 unimplemented_feature_nlines_dict = dict()
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	331
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	332 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	333 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	334 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	335 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	336 # 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	337 continue
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	338 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	339 # 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	340 continue
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	341 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	342 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	343 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	344 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	345 try:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	346 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	347 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	348 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	349 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	350 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	351 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	352 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	353 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	354 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	355 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	356 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	357 add_cds_to_dict(cols, cds_parent_dict)
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	358 elif feature_type in unimplemented_feature_nlines_dict:
b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	359 unimplemented_feature_nlines_dict[feature_type] += 1
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	360 else:
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	361 unimplemented_feature_nlines_dict[feature_type] = 0
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	362 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	363 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	364
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	365 for unimplemented_feature, nlines in unimplemented_feature_nlines_dict.items():
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	366 print("Skipped %d lines in GFF3 file '%s': '%s' is not an implemented feature type" % (nlines, filename, unimplemented_feature), file=sys.stderr)
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	367
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	368 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	369 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	370
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	371 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	372 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	373 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	374
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	375 # Read the FASTA files a first time to:
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	376 # - determine for each file if we need to force the removal of the version
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	377 # from the transcript id
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	378 # - fill gene_transcripts_dict when keeping only the canonical transcripts
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	379 force_remove_id_version_file_list = []
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	380 gene_transcripts_dict = dict()
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	381 for fasta_arg in options.fasta:
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	382 force_remove_id_version = False
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	383 found_gene_transcript = False
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	384 for entry in FASTAReader_gen(fasta_arg):
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	385 # Extract the transcript id by removing everything after the first space and then removing the version if needed
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	386 transcript_id = remove_id_version(entry.header[1:].lstrip().split(' ')[0], force_remove_id_version)
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	387
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	388 transcript = fetch_transcript_and_gene(conn, transcript_id)
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	389 if not transcript and not found_gene_transcript:
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	390 # We have not found a proper gene transcript in this file yet,
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	391 # try to force the removal of the version from the transcript id
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	392 transcript_id = remove_id_version(entry.header[1:].lstrip().split(' ')[0], True)
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	393 transcript = fetch_transcript_and_gene(conn, transcript_id)
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	394 # Remember that we need to force the removal for this file
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	395 if transcript:
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	396 force_remove_id_version = True
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	397 force_remove_id_version_file_list.append(fasta_arg)
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	398 print("Forcing removal of id version in FASTA file '%s'" % fasta_arg, file=sys.stderr)
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	399 if not transcript:
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	400 print("Transcript '%s' in FASTA file '%s' not found in the gene feature information" % (transcript_id, fasta_arg), file=sys.stderr)
f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	401 continue
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	402 if options.filter != 'canonical':
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	403 break
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	404 found_gene_transcript = True
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	405
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	406 if len(entry.sequence) % 3 != 0:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	407 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	408 transcript_biotype = transcript['biotype'] # This is the biotype of the transcript or, if that is NULL, the one of the gene
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	409 if transcript_biotype and transcript_biotype != 'protein_coding':
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	410 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	411 gene_transcripts_dict.setdefault(transcript['gene_id'], []).append((transcript_id, transcript['is_canonical'], len(entry.sequence)))
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	412
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	413 if options.filter == 'canonical':
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	414 selected_transcript_ids = []
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	415 for gene_id, transcript_tuples in gene_transcripts_dict.items():
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	416 canonical_transcript_ids = [id_ for (id_, is_canonical, _) in transcript_tuples if is_canonical]
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	417 if not canonical_transcript_ids:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	418 # Select the transcript with the longest sequence. If more than
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	419 # one transcripts have the same longest sequence for a gene, the
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	420 # first one to appear in the FASTA file is selected.
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	421 selected_transcript_id = max(transcript_tuples, key=lambda transcript_tuple: transcript_tuple[2])[0]
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	422 elif len(canonical_transcript_ids) > 1:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	423 raise Exception("Gene %s has more than 1 canonical transcripts" % (gene_id))
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	424 else:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	425 selected_transcript_id = canonical_transcript_ids[0]
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	426 selected_transcript_ids.append(selected_transcript_id)
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	427
6 56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	428 regions = [_.strip().lower() for _ in options.regions.split(",")]
56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	429 with open(options.of, 'w') as output_fasta_file, open(options.ff, 'w') as filtered_fasta_file:
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	430 for fasta_arg in options.fasta:
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	431 force_remove_id_version = fasta_arg in force_remove_id_version_file_list
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	432 for entry in FASTAReader_gen(fasta_arg):
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	433 transcript_id = remove_id_version(entry.header[1:].lstrip().split(' ')[0], force_remove_id_version)
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	434
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	435 transcript = fetch_transcript_and_gene(conn, transcript_id)
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	436 if not transcript:
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	437 print("Transcript '%s' in FASTA file '%s' not found in the gene feature information" % (transcript_id, fasta_arg), file=sys.stderr)
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	438 continue
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	439
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	440 if options.filter == 'canonical':
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	441 # We already filtered out non-protein-coding transcripts when populating gene_transcripts_dict
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	442 if transcript_id not in selected_transcript_ids:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	443 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	444 elif options.filter == 'coding':
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	445 if len(entry.sequence) % 3 != 0:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	446 print("Transcript '%s' in FASTA file '%s' has a coding sequence length which is not multiple of 3, removing from FASTA output" % (transcript_id, fasta_arg), file=sys.stderr)
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	447 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	448 transcript_biotype = transcript['biotype'] # This is the biotype of the transcript or, if that is NULL, the one of the gene
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	449 if transcript_biotype and transcript_biotype != 'protein_coding':
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	450 print("Transcript %s has biotype %s (not protein-coding), removing from FASTA output" % (transcript_id, transcript_biotype), file=sys.stderr)
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	451 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	452
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	453 if options.headers == "TranscriptId_species":
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	454 # Change the FASTA header to '>TranscriptId_species', as required by TreeBest
284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	455 # Remove any underscore in the species
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	456 entry.header = ">%s_%s" % (transcript_id, transcript['species'].replace('_', ''))
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	457 elif options.headers == "GeneSymbol-TranscriptID_species":
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	458 # Remove any underscore in the species
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	459 entry.header = ">%s-%s_%s" % (transcript['gene_symbol'], transcript_id, transcript['species'].replace('_', ''))
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	460 elif options.headers == "TranscriptSymbol-TranscriptID_species":
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	461 # Remove any underscore in the species
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	462 entry.header = ">%s-%s_%s" % (transcript['transcript_symbol'], transcript_id, transcript['species'].replace('_', ''))
4 284f64ad9d43 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit cda3ecab1a34376cc7d4d392a34dc810847cbf0b-dirty earlhaminst parents: 3 diff changeset	463
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	464 if transcript['seq_region_name'].lower() in regions:
8 92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	465 entry.print(filtered_fasta_file)
6 56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	466 else:
8 92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	467 entry.print(output_fasta_file)
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	468
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	469 conn.close()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	470
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	471
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	472 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	473 __main__()

Mercurial > repos > earlhaminst > gstf_preparation

annotate gstf_preparation.py @ 11:dbe37a658cd2 draft