annotate filter_utr.py @ 22:4309798bfb61 draft default tip

planemo upload for repository https://github.com/mvdbeek/dapars commit b1b007c561ea6c9db145c88b6b128d66ecd05e24-dirty
author mvdbeek
date Fri, 30 Oct 2015 12:30:22 -0400
parents bb84ee2f2137
children
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
rev   line source
0
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
1 #!/usr/bin/env python
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
2 # Filter out UTRs and, if a gene has multiple UTRs, return a single UTR with minimum start and maximum end coordinate
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
3 # usage: python filter_utr.py input.gtf output.gtf
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
4
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
5 import sys
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
6 from collections import OrderedDict
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
7
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
8 def get_gtf_fields():
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
9 return [ "chr", "source", "feature", "start", "end", "score", "strand", "frame", "group" ]
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
10
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
11 def get_feature_dict(line, gtf_fields, utr_dict, feature="UTR"):
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
12 """
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
13 Return a dictionary with lines of a GTF if the line describes a UTR.
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
14 Key is the first attribute of the group.
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
15 """
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
16 if line.split("\t")[2] == feature:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
17 fields = line.strip().split("\t")
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
18 fields[3] = int(fields[3])
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
19 fields[4] = int(fields[4])
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
20 gene_id = fields[-1].split("\"")[1]
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
21 if not gene_id in utr_dict:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
22 utr_dict[gene_id] = [OrderedDict(zip(gtf_fields, fields))]
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
23 else:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
24 utr_dict[gene_id].append(OrderedDict(zip(gtf_fields, fields)))
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
25
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
26 def remove_five_prime_utrs(utr_dict, start_codon_dict):
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
27 for gene, features in start_codon_dict.iteritems():
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
28 is_reverse = features[0]["strand"] == "-"
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
29 if is_reverse:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
30 stop_codon_end = min([feature["start"] for feature in features])
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
31 else:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
32 stop_codon_end = max([feature["end"] for feature in features])# get last start codon
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
33 to_remove = []
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
34 if gene in utr_dict:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
35 for utr in utr_dict[gene]:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
36 start = utr["start"]
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
37 end = utr["end"]
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
38 if is_reverse:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
39 if end >= stop_codon_end:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
40 to_remove.append(utr)
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
41 else:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
42 if start <= stop_codon_end:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
43 to_remove.append(utr)
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
44 [utr_dict[gene].remove(utr) for utr in to_remove ]
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
45
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
46
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
47 def get_longest_utr(utr_dict):
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
48 """
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
49 Start of the composite utr is the most 5p start, end is the most 3p end.
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
50 """
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
51 gtf = []
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
52 for gene_id, values in utr_dict.iteritems():
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
53 if not values:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
54 continue
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
55 if len(values) == 1:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
56 values[0]["start"] = str(values[0]["start"])
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
57 values[0]["end"] = str(values[0]["end"])
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
58 gtf.append( "\t".join( values[0].values() ) )
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
59 else:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
60 start = min( [fields["start"] for fields in values] )
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
61 end = max( [fields["end"] for fields in values] )
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
62 values[0]["start"] = str(start)
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
63 values[0]["end"] = str(end)
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
64 gtf.append( "\t".join( values[0].values() ) )
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
65 return gtf
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
66
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
67 def main():
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
68 utr_dict = OrderedDict()
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
69 start_codon_dict = {}
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
70 header = []
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
71 gtf_fields = get_gtf_fields()
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
72 with open(sys.argv[1]) as input:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
73 for line in input:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
74 if line.startswith("#"):
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
75 header.append( line.strip() )
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
76 else:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
77 get_feature_dict(line, gtf_fields, utr_dict, feature="UTR")
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
78 get_feature_dict(line, gtf_fields, start_codon_dict, feature="CDS")
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
79 remove_five_prime_utrs(utr_dict, start_codon_dict)
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
80 gtf = header + get_longest_utr(utr_dict)
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
81 if len(sys.argv) == 3:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
82 with open(sys.argv[2], "w") as output:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
83 [output.write(line + "\n") for line in gtf]
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
84 else:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
85 for line in gtf:
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
86 print(line)
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
87
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
88 if __name__ == "__main__":
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
89 main()
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
90
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
91
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
92
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
93
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
94
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
95
bb84ee2f2137 planemo upload for repository https://github.com/mvdbeek/dapars commit 868f8f2f7ac5d70c39b7d725ff087833b0f24f52-dirty
mvdbeek
parents:
diff changeset
96