gstf_preparation: gstf_preparation.py annotate

annotate gstf_preparation.py @ 15:9c62ad7dd113 draft default tip

"planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb"

author	earlhaminst
date	Thu, 29 Apr 2021 19:46:38 +0000
parents	598e9172b8e7
children

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 import json
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	2 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	3 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	4 import sqlite3
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	5 import sys
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	6
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	7 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	8 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	9
8 92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	10
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	11 def asbool(val):
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	12 if isinstance(val, str):
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	13 val_lower = val.strip().lower()
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	14 if val_lower in ('true', '1'):
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	15 return True
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	16 elif val_lower in ('false', '0'):
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	17 return False
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	18 else:
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	19 raise ValueError(f"Cannot convert {val} to bool")
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	20 else:
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	21 return bool(val)
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	22
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	23
51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	24 class Sequence:
8 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 __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	26 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	27 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	28 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	29
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	30 @property
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	31 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	32 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	33 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	34 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	35
92f3966d5bc3 planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 88ba62ae8c3d9587a0015c72209242ad0c1df0c2 earlhaminst parents: 6 diff changeset	36 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	37 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	38 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	39 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	40
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	41
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	42 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	43 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	44 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	45 while True:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	46 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	47 return
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	48 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	49 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	50 sequence_parts = []
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	51 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	52 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	53 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	54 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	55 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	56
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	57
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	58 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	59 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	60
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	61 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	62 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	63
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	64 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	65 (version, ))
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	66
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	67 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	68 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	69 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	70 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	71 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	72 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	73 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	74 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	75 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	76 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	77 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	78
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	79 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	80 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	81 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	82 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	83 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	84 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	85 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	86 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	87
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	88 # 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	89 # 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	90 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	91 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	92 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	93 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	94
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	95 conn.commit()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	96
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	97
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	98 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	99 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	100
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	101 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	102 (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	103 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	104
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	105
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	106 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	107 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	108
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	109
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	110 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	111 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	112 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	113 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	114 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	115 return id_
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	116
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	117
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	118 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	119 d = {
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	120 '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	121 '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	122 }
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	123 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	124 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	125 (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	126 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	127 tag = 'id'
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	128 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	129 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	130 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	131 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	132 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	133 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	134 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	135 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	136 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	137 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	138 else:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	139 raise Exception(f"Unrecognized strand: {cols[6]}")
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	140 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	141 # 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	142 # 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	143 # 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	144 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	145 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	146 return d
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	147
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	148
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	149 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	150 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	151 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	152 if not gene['id']:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	153 raise Exception(f"Id not found among column 9 attribute tags: {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	154 gene.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	155 '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	156 '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	157 '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	158 'species': species,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	159 '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	160 '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	161 })
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	162 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	163 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	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_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	167 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	168 transcript.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	169 '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	170 '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	171 '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	172 '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	173 })
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	174 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	175
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 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	178 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	179 exon.update({
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	180 '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	181 '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	182 '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	183 'species': species,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	184 })
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	185 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	186 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	187
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	188
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	189 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	190 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	191 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	192 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	193 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	194 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	195 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	196
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	197
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	198 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	199 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	200 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	201 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	202 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	203
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	204 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	205 translation = {
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	206 'CDS': [],
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	207 'id': None,
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	208 '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	209 '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	210 '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	211 '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	212 }
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	213 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	214 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	215 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	216 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	217 cds_list = cds_parent_dict[transcript_id]
14 598e9172b8e7 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit ea67c766934266e690d05e3f9ebb4cca12b8e3e7" earlhaminst parents: 13 diff changeset	218 unique_cds_ids = {cds['id'] for cds in cds_list}
598e9172b8e7 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit ea67c766934266e690d05e3f9ebb4cca12b8e3e7" earlhaminst parents: 13 diff changeset	219 if len(unique_cds_ids) > 1:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	220 msg = f"""Found multiple CDS IDs ({unique_cds_ids}) for transcript '{transcript_id}'.
14 598e9172b8e7 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit ea67c766934266e690d05e3f9ebb4cca12b8e3e7" earlhaminst parents: 13 diff changeset	221 This is not supported by the Ensembl JSON format. If a CDS is split across
598e9172b8e7 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit ea67c766934266e690d05e3f9ebb4cca12b8e3e7" earlhaminst parents: 13 diff changeset	222 multiple discontinuous genomic locations, the GFF3 standard requires that all
598e9172b8e7 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit ea67c766934266e690d05e3f9ebb4cca12b8e3e7" earlhaminst parents: 13 diff changeset	223 corresponding lines use the same ID attribute."""
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	224 raise Exception(msg)
14 598e9172b8e7 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit ea67c766934266e690d05e3f9ebb4cca12b8e3e7" earlhaminst parents: 13 diff changeset	225 cds_id = unique_cds_ids.pop()
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	226 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	227 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	228 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	229 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	230 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	231 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	232 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	233 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	234 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	235 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	236 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	237 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	238 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	239 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	240 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	241 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	242 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	243 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	244 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	245 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	246 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	247 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	248 if derived_translation_start > translation['start']:
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	249 raise Exception(f"Transcript {transcript_id} has the start of CDS {cds_id} overlapping with the UTR end")
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	250 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	251 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	252 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	253 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	254 if derived_translation_end < translation['end']:
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	255 raise Exception(f"Transcript {transcript_id} has the end of CDS {cds_id} overlapping with the UTR start")
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	256 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	257 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	258 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	259 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	260
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	261 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	262 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	263 # 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	264 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	265 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	266 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	267
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	268
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	269 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	270 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	271
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	272 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	273 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	274 # 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	275 continue
12 99bae410128c "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 4579d0c461c30183a3092d84013e30f53f072ca1-dirty" earlhaminst parents: 11 diff changeset	276 if 'confidence' in gene and gene['confidence'].lower() != 'high':
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	277 print("Gene {} has confidence {} (not high), discarding".format(gene['id'], gene['confidence']), 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	278 continue
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	279 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	280 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	281 (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	282
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	283 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	284 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	285 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	286 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	287 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	288 biotype = transcript.get('biotype')
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	289 is_canonical = asbool(transcript.get('is_canonical', False))
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	290 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	291 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	292 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	293 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	294 except Exception as e:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	295 raise Exception(f"Error while inserting {to_insert} into transcript table: {e}")
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	296
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	297 conn.commit()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	298
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	299
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	300 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	301 """
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	302 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	303 `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	304 """
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	305 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	306 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	307 else:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	308 return s
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	309
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	310
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	311 def __main__():
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	312 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	313 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	314 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	315 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	316 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	317 parser.add_option('--headers', type='choice',
12 99bae410128c "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 4579d0c461c30183a3092d84013e30f53f072ca1-dirty" earlhaminst parents: 11 diff changeset	318 choices=['TranscriptId_species', 'TranscriptID-GeneSymbol_species', 'TranscriptID-TranscriptSymbol_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	319 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	320 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	321 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	322 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	323 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	324
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	325 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	326 if args:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	327 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	328
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	329 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	330 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	331 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	332 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	333
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	334 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	335 try:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	336 (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	337 except ValueError:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	338 raise Exception(f"Argument for --gff3 '{gff3_arg}' is not in the SPECIES:FILENAME format")
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	339 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	340 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	341 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	342 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	343 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	344 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	345 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	346
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	347 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	348 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	349 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	350 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	351 # 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	352 continue
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	353 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	354 # 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	355 continue
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	356 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	357 if len(cols) != 9:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	358 raise Exception(f"Line {i} in file '{filename}': '{line}' does not have 9 columns")
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	359 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	360 try:
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	361 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	362 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	363 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	364 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	365 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	366 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	367 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	368 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	369 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	370 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	371 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	372 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	373 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	374 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	375 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	376 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	377 except Exception as e:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	378 print(f"Line {i} in file '{filename}': {e}", 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	379
5 b3ba0c84667c planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 95bab1105cf8a7b07c668f08f712399e8775a4ae earlhaminst parents: 4 diff changeset	380 for unimplemented_feature, nlines in unimplemented_feature_nlines_dict.items():
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	381 print(f"Skipped {nlines} lines in GFF3 file '{filename}': '{unimplemented_feature}' is not an implemented feature type", 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	382
0 28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	383 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	384 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	385
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	386 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	387 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	388 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	389
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 # 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	391 # - 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	392 # 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	393 # - 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	394 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	395 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	396 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	397 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	398 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	399 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	400 # 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	401 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	402
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	403 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	404 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	405 # 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	406 # 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	407 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	408 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	409 # 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	410 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	411 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	412 force_remove_id_version_file_list.append(fasta_arg)
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	413 print(f"Forcing removal of id version in FASTA file '{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	414 if not transcript:
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	415 print(f"Transcript '{transcript_id}' in FASTA file '{fasta_arg}' not found in the gene feature information", file=sys.stderr)
9 f4acbfe8d6fe planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 2f56285b1ef694d732c8b2637e3e924f8a626e55 earlhaminst parents: 8 diff changeset	416 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	417 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	418 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	419 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	420
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	421 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	422 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	423 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	424 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	425 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	426 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	427
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	428 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	429 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	430 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	431 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	432 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	433 # 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	434 # 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	435 # 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	436 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	437 elif len(canonical_transcript_ids) > 1:
15 9c62ad7dd113 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit a4e49466bf746633ddc26d052b65ca41576d18fb" earlhaminst parents: 14 diff changeset	438 raise Exception(f"Gene {gene_id} has more than 1 canonical transcripts")
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	439 else:
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	440 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	441 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	442
6 56bbdbfe3eaa planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit fa875eea77a9471acada2b7b8882a0467994c960 earlhaminst parents: 5 diff changeset	443 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	444 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	445 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	446 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	447 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	448 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	449
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	450 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	451 if not transcript:
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	452 print(f"Transcript '{transcript_id}' in FASTA file '{fasta_arg}' not found in the gene feature information", 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	453 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	454
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	455 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	456 # 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	457 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	458 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	459 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	460 if len(entry.sequence) % 3 != 0:
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	461 print(f"Transcript '{transcript_id}' in FASTA file '{fasta_arg}' has a coding sequence length which is not multiple of 3, removing from FASTA output", 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	462 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	463 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	464 if transcript_biotype and transcript_biotype != 'protein_coding':
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	465 print(f"Transcript {transcript_id} has biotype {transcript_biotype} (not protein-coding), removing from FASTA output", 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	466 continue
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	467
dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	468 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	469 # 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	470 # Remove any underscore in the species
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	471 entry.header = ">{}_{}".format(transcript_id, transcript['species'].replace('_', ''))
12 99bae410128c "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 4579d0c461c30183a3092d84013e30f53f072ca1-dirty" earlhaminst parents: 11 diff changeset	472 elif options.headers == "TranscriptID-GeneSymbol_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	473 # Remove any underscore in the species
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	474 entry.header = ">{}-{}_{}".format(transcript_id, transcript['gene_symbol'], transcript['species'].replace('_', ''))
12 99bae410128c "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 4579d0c461c30183a3092d84013e30f53f072ca1-dirty" earlhaminst parents: 11 diff changeset	475 elif options.headers == "TranscriptID-TranscriptSymbol_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	476 # Remove any underscore in the species
13 51a7a2a82902 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 9178f870760132962f8d3a26ea55c201880bb018-dirty" earlhaminst parents: 12 diff changeset	477 entry.header = ">{}-{}_{}".format(transcript_id, transcript['transcript_symbol'], 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	478
11 dbe37a658cd2 "planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 133a11e7195f8da83c5b661d8babb3f6d9e18812" earlhaminst parents: 10 diff changeset	479 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	480 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	481 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	482 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	483
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	484 conn.close()
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	485
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	486
28879ca33b5f planemo upload for repository https://github.com/TGAC/earlham-galaxytools/tree/master/tools/gstf_preparation commit 651fae48371f845578753052c6fe173e3bb35670 earlhaminst parents: diff changeset	487 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	488 __main__()

Mercurial > repos > earlhaminst > gstf_preparation

annotate gstf_preparation.py @ 15:9c62ad7dd113 draft default tip