Mercurial > repos > cpt > cpt_gff_to_gbk

annotate gff2gb.py @ 4:35a6c466e270 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
planemo upload commit 852ac96ca53a2ffa0947e6df5e24671866b642f5
author cpt
date Sun, 23 Jul 2023 01:38:52 +0000
parents c8fcb7246ac3
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
3
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
1 #!/usr/bin/env python
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
2 """Convert a GFF and associated FASTA file into GenBank format.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
3
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
4 Usage:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
5 gff_to_genbank.py <GFF annotation file> <FASTA sequence file>
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
6 """
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
7 import argparse
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
8 import sys
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
9 import re
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
10 import copy
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
11 import itertools
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
12 import logging
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
13 from Bio import SeqIO
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
14
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
15 # from Bio.Alphabet import generic_dna
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
16 from Bio.SeqFeature import CompoundLocation, FeatureLocation
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
17 from CPT_GFFParser import gffParse, gffWrite
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
18 from gff3 import (
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
19 feature_lambda,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
20 wa_unified_product_name,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
21 is_uuid,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
22 feature_test_type,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
23 fsort,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
24 feature_test_true,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
25 feature_test_quals,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
26 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
27
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
28 default_name = re.compile(r"^gene_(\d+)$")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
29 logging.basicConfig(level=logging.INFO)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
30
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
31
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
32 def rename_key(ds, k_f, k_t):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
33 """Rename a key in a dictionary and return it, FP style"""
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
34 # If they key is not in the dictionary, just return immediately
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
35 if k_f not in ds:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
36 return ds
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
37
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
38 # Otherwise, we check if the target key is in there
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
39 if k_t in ds:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
40 # If it is, we need to append
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
41 ds[k_t] += ds[k_f]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
42 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
43 # if not, we can just set.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
44 ds[k_t] = ds[k_f]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
45
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
46 # Remove source
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
47 del ds[k_f]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
48 return ds
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
49
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
50
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
51 def gff3_to_genbank(gff_file, fasta_file, transltbl):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
52 fasta_input = SeqIO.to_dict(SeqIO.parse(fasta_file, "fasta")) # , generic_dna))
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
53 gff_iter = gffParse(gff_file, fasta_input)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
54
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
55 for record in gff_iter:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
56 yield handle_record(record, transltbl)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
57
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
58
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
59 def handle_non_gene_features(features):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
60 # These are NON-GENE features (maybe terminators? etc?)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
61 for feature in feature_lambda(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
62 features,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
63 feature_test_type,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
64 {"type": "gene"},
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
65 subfeatures=False,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
66 invert=True,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
67 recurse=True, # used to catch RBS from new apollo runs (used to be False)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
68 ):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
69 if feature.type in (
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
70 "terminator",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
71 "tRNA",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
72 "Shine_Dalgarno_sequence",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
73 "sequence_feature",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
74 "recombination_feature",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
75 "sequence_alteration",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
76 "binding_site",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
77 ):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
78 yield feature
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
79 elif feature.type in ("CDS",):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
80 pass
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
81 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
82 yield feature
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
83
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
84
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
85 def fminmax(feature):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
86 fmin = None
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
87 fmax = None
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
88 for sf in feature_lambda([feature], feature_test_true, {}, subfeatures=True):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
89 if fmin is None:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
90 fmin = sf.location.start
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
91 fmax = sf.location.end
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
92 if sf.location.start < fmin:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
93 fmin = sf.location.start
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
94 if sf.location.end > fmax:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
95 fmax = sf.location.end
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
96 return fmin, fmax
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
97
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
98
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
99 def fix_gene_boundaries(feature):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
100 # There is a frustrating bug in apollo whereby we have created gene
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
101 # features which are LARGER than expected, but we cannot see this.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
102 # We only see a perfect sized gene + great SD together.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
103 #
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
104 # So, we have this awful hack to clamp the location of the gene
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
105 # feature to the contained mRNAs. This is good enough for now.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
106 fmin, fmax = fminmax(feature)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
107 if feature.location.strand > 0:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
108 feature.location = FeatureLocation(fmin, fmax, strand=1)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
109 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
110 feature.location = FeatureLocation(fmin, fmax, strand=-1)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
111 return feature
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
112
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
113
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
114 def fix_gene_qualifiers(name, feature, fid):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
115 for mRNA in feature.sub_features:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
116 mRNA.qualifiers["locus_tag"] = "CPT_%s_%03d" % (name, fid)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
117 # And some exons below that
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
118 sf_replacement = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
119 for sf in mRNA.sub_features:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
120 # We set a locus_tag on ALL sub features
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
121 sf.qualifiers["locus_tag"] = "CPT_%s_%03d" % (name, fid)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
122 # Remove Names which are UUIDs
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
123 # NOT GOOD PRACTICE
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
124 try:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
125 if is_uuid(sf.qualifiers["Name"][0]):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
126 del sf.qualifiers["Name"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
127 except KeyError:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
128 continue # might should go back to pass, I have not put thought into this still
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
129
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
130 # If it is the RBS exon (mis-labelled by apollo as 'exon')
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
131 if sf.type == "exon" and len(sf) < 10:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
132 sf.type = "Shine_Dalgarno_sequence"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
133 sf_replacement.append(sf)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
134 # and if it is the CDS
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
135 elif sf.type == "CDS":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
136 # Update CDS qualifiers with all info that was on parent
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
137 sf.qualifiers.update(feature.qualifiers)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
138 sf_replacement.append(sf)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
139 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
140 sf_replacement.append(sf)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
141 if mRNA.type == "tRNA":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
142 mRNA.qualifiers["product"] = mRNA.qualifiers["Name"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
143 # Handle multiple child CDS features by merging them.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
144 # Replace the subfeatures on the mRNA
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
145 mRNA.sub_features = merge_multi_cds(sf_replacement)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
146 return feature
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
147
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
148
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
149 def fix_frameshifted(features):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
150 logging.info("Fixing Frameshifted group: [%s]", str(features))
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
151 genes = features
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
152 # Find all mRNAs (plus reduce nested list into flattened one)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
153 mRNAs = sum([f.sub_features for f in genes], [])
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
154 # Find all CDSs (plus reduce nested list into flattened one)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
155 cdss = sum([m.sub_features for m in mRNAs], [])
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
156 # List to store the RBSs which we'll break apart + re-attach later.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
157 rbss = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
158 # List to store all of the CDSs (i.e. cdss - rbss)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
159 cdss2 = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
160 # Copy genes + clean out subfeatures. We'll re-use these constructs.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
161 fixed_features = copy.deepcopy(genes)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
162 for f in fixed_features:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
163 f.sub_features = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
164 # Copy / empty out mRNAs
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
165 fixed_mrnas = copy.deepcopy(mRNAs)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
166 for f in fixed_mrnas:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
167 f.sub_features = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
168 f.qualifiers = {}
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
169 # Fill rbss + cdss2
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
170 for cds in cdss:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
171 if "frameshift" in cds.qualifiers:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
172 del cds.qualifiers["frameshift"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
173 # Ignore short features, as those are RBSs
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
174 if len(cds) < 15:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
175 rbss.append(cds)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
176 continue
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
177 # Otherwise cdss.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
178 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
179 cdss2.append(cds)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
180 # Ok, now have cdss2 to deal with.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
181 other = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
182 # Find the two with least value for distance between end / start (strand aware).
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
183 # For every possible pair, we'll check their distance
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
184 match_data = {}
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
185 for (a, b) in itertools.permutations(cdss2, 2):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
186 if a.location.start < b.location.start:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
187 # A is downstream of B
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
188 match_data[(a, b)] = b.location.start - a.location.end
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
189 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
190 match_data[(a, b)] = a.location.start - b.location.end
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
191
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
192 # Now we'll find the features which are closest in terms of start/end
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
193 ((merge_a, merge_b), value) = max(match_data.items(), key=lambda kv: kv[1])
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
194 # And get the non-matching features into other
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
195 for f in cdss2:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
196 if f != merge_a and f != merge_b:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
197 other.append(f)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
198 # Back to the merge_a/b
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
199 # With those, we'll merge them into one feature, and discard the other.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
200 merge_a.location = CompoundLocation([merge_a.location, merge_b.location])
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
201 # The gene + RBSs should be identical and two/two.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
202 assert len(fixed_features) == 2
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
203 # If not, we can just duplicate the RBS, doesn't matter.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
204 noRBS = len(rbss) == 0
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
205 if len(rbss) != 2 and not noRBS:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
206 rbss = [rbss[0], copy.deepcopy(rbss[0])]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
207 # Now re-construct.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
208 gene_0 = fixed_features[0]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
209 gene_1 = fixed_features[1]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
210 mRNA_0 = fixed_mrnas[0]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
211 mRNA_1 = fixed_mrnas[1]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
212
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
213 if not noRBS:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
214 mRNA_0.sub_features = [rbss[0], merge_a]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
215 mRNA_1.sub_features = other + [rbss[1]]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
216 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
217 mRNA_0.sub_features = [merge_a]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
218 mRNA_1.sub_features = other
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
219 mRNA_0 = fix_gene_boundaries(mRNA_0)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
220 mRNA_1 = fix_gene_boundaries(mRNA_1)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
221
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
222 gene_0.sub_features = [mRNA_0]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
223 gene_1.sub_features = [mRNA_1]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
224 gene_0 = fix_gene_boundaries(gene_0)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
225 gene_1 = fix_gene_boundaries(gene_1)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
226
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
227 return fixed_features
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
228
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
229
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
230 def fix_frameshifts(features):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
231 # Collect all gene features where at least one subfeature has a
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
232 # frameshift=??? annotation.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
233 def has_frameshift_qual(f):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
234 return (
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
235 len(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
236 list(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
237 feature_lambda(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
238 f.sub_features, feature_test_quals, {"frameshift": None}
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
239 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
240 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
241 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
242 > 0
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
243 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
244
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
245 def has_frameshift_qual_val(f, val):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
246 return (
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
247 len(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
248 list(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
249 feature_lambda(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
250 f.sub_features, feature_test_quals, {"frameshift": val}
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
251 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
252 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
253 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
254 > 0
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
255 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
256
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
257 def get_frameshift_qual(f):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
258 for f in feature_lambda(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
259 f.sub_features, feature_test_quals, {"frameshift": None}
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
260 ):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
261 return f.qualifiers["frameshift"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
262
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
263 to_frameshift = [x for x in features if x.type == "gene" and has_frameshift_qual(x)]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
264 fixed = [x for x in features if x not in to_frameshift]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
265
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
266 frameshift_keys = set(sum(map(get_frameshift_qual, to_frameshift), []))
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
267 for key in frameshift_keys:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
268 # Get features matching that key
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
269 current = [x for x in to_frameshift if has_frameshift_qual_val(x, key)]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
270 # Fix them and append them
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
271 fixed += fix_frameshifted(current)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
272
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
273 return fixed
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
274
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
275
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
276 def remove_useless_features(features):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
277 # Drop mRNAs, apollo crap, useless CDSs
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
278 for f in features:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
279 if f.type in (
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
280 "non_canonical_three_prime_splice_site",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
281 "non_canonical_five_prime_splice_site",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
282 "stop_codon_read_through",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
283 "mRNA",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
284 "exon",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
285 ):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
286 continue
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
287 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
288 if f.type == "CDS" and len(f) < 10:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
289 # Another RBS mistake
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
290 continue
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
291 # We use the full GO term, but it should be less than that.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
292 if f.type == "Shine_Dalgarno_sequence":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
293 f.type = "RBS"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
294
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
295 if f.type == "sequence_feature":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
296 f.type = "misc_feature"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
297
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
298 if f.type == "recombination_feature":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
299 f.type = "misc_recomb"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
300
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
301 if f.type == "sequence_alteration":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
302 f.type = "variation"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
303
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
304 if f.type == "binding_site":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
305 f.type = "misc_binding"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
306
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
307 yield f
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
308
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
309
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
310 def merge_multi_cds(mRNA_sf):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
311 cdss = [x for x in mRNA_sf if x.type == "CDS"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
312 non_cdss = [x for x in mRNA_sf if x.type != "CDS"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
313 if len(cdss) <= 1:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
314 return non_cdss + cdss
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
315 else:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
316 # Grab all locations, and sort them so we can work with them rationally.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
317 locations = sorted([x.location for x in cdss], key=lambda y: y.start)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
318 # Pick randomly a main CDS
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
319 main_cds = cdss[0]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
320 # We'll merge the other CDSs into this one.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
321 main_cds.location = CompoundLocation(locations)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
322 return non_cdss + [main_cds]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
323
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
324
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
325 def handle_record(record, transltbl):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
326 full_feats = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
327 for feature in fsort(record.features):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
328 if (
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
329 feature.type == "region"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
330 and "source" in feature.qualifiers
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
331 and "GenBank" in feature.qualifiers["source"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
332 ):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
333 feature.type = "source"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
334
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
335 if "comment1" in feature.qualifiers:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
336 del feature.qualifiers["comment1"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
337
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
338 if "Note" in feature.qualifiers:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
339 record.annotations = feature.qualifiers
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
340 if len(feature.qualifiers["Note"]) > 1:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
341 record.annotations["comment"] = feature.qualifiers["Note"][1]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
342 del feature.qualifiers["Note"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
343
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
344 if "comment" in feature.qualifiers:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
345 del feature.qualifiers["comment"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
346
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
347 # We'll work on a separate copy of features to avoid modifying a list
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
348 # we're iterating over
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
349 replacement_feats = []
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
350 replacement_feats += list(handle_non_gene_features(record.features))
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
351
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
352 # Renumbering requires sorting
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
353 fid = 0
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
354 for feature in fsort(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
355 feature_lambda(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
356 record.features, feature_test_type, {"type": "gene"}, subfeatures=True
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
357 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
358 ):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
359 # Our modifications only involve genes
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
360 fid += 1
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
361
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
362 feature = fix_gene_boundaries(feature)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
363 # Which have mRNAs we'll drop later
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
364 feature = fix_gene_qualifiers(record.id, feature, fid)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
365
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
366 # Wipe out the parent gene's data, leaving only a locus_tag
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
367 feature.qualifiers = {"locus_tag": "CPT_%s_%03d" % (record.id, fid)}
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
368
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
369 # Patch our features back in (even if they're non-gene features)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
370 replacement_feats.append(feature)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
371
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
372 replacement_feats = fix_frameshifts(replacement_feats)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
373 # exit(0)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
374 flat_features = feature_lambda(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
375 replacement_feats, lambda x: True, {}, subfeatures=True
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
376 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
377
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
378 flat_features = remove_useless_features(flat_features)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
379
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
380 # Meat of our modifications
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
381 for flat_feat in flat_features:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
382 # Try and figure out a name. We gave conflicting instructions, so
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
383 # this isn't as trivial as it should be.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
384 protein_product = wa_unified_product_name(flat_feat)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
385
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
386 for x in (
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
387 "source",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
388 "phase",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
389 "Parent",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
390 "ID",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
391 "owner",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
392 "date_creation",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
393 "date_last_modified",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
394 "datasetSource",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
395 ):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
396 if x in flat_feat.qualifiers:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
397 if x == "ID":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
398 flat_feat._ID = flat_feat.qualifiers["ID"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
399 del flat_feat.qualifiers[x]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
400
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
401 # Add product tag
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
402 if flat_feat.type == "CDS":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
403 flat_feat.qualifiers["product"] = [protein_product]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
404 flat_feat.qualifiers["transl_table"] = [transltbl]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
405 if "Product" in flat_feat.qualifiers:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
406 del flat_feat.qualifiers["Product"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
407 elif flat_feat.type == "RBS":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
408 if "locus_tag" not in flat_feat.qualifiers.keys():
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
409 continue
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
410
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
411 elif flat_feat.type == "terminator":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
412 flat_feat.type = "regulatory"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
413 flat_feat.qualifiers = {"regulatory_class": "terminator"}
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
414
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
415 # In genbank format, note is lower case.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
416 flat_feat.qualifiers = rename_key(flat_feat.qualifiers, "Note", "note")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
417 flat_feat.qualifiers = rename_key(flat_feat.qualifiers, "description", "note")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
418 flat_feat.qualifiers = rename_key(flat_feat.qualifiers, "protein", "note")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
419 flat_feat.qualifiers = rename_key(flat_feat.qualifiers, "Dbxref", "db_xref")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
420 if "Name" in flat_feat.qualifiers:
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
421 del flat_feat.qualifiers["Name"]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
422
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
423 # more apollo nonsense
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
424 if "Manually set translation start" in flat_feat.qualifiers.get("note", []):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
425 flat_feat.qualifiers["note"].remove("Manually set translation start")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
426
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
427 # Append the feature
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
428 full_feats.append(flat_feat)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
429
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
430 # Update our features
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
431 record.features = fsort(full_feats)
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
432 # Strip off record names that would cause crashes.
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
433 record.name = record.name[0:16]
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
434 return record
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
435
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
436
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
437 if __name__ == "__main__":
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
438 # Grab all of the filters from our plugin loader
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
439 parser = argparse.ArgumentParser(description="Convert gff3 to gbk")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
440 parser.add_argument("gff_file", type=argparse.FileType("r"), help="GFF3 file")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
441 parser.add_argument("fasta_file", type=argparse.FileType("r"), help="Fasta Input")
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
442 parser.add_argument(
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
443 "--transltbl",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
444 type=int,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
445 default=11,
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
446 help="Translation Table choice for CDS tag, default 11",
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
447 )
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
448 args = parser.parse_args()
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
449
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
450 for record in gff3_to_genbank(**vars(args)):
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
451 record.annotations["molecule_type"] = "DNA"
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
452 # record.seq.alphabet = generic_dna
c8fcb7246ac3 planemo upload commit 94b0cd1fff0826c6db3e7dc0c91c0c5a8be8bb0c
cpt
parents:
diff changeset
453 SeqIO.write([record], sys.stdout, "genbank")