BioDSL 1.0.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/.gitignore +10 -0
- data/BioDSL.gemspec +64 -0
- data/LICENSE +339 -0
- data/README.md +205 -0
- data/Rakefile +94 -0
- data/examples/fastq_to_fasta.rb +8 -0
- data/lib/BioDSL/cary.rb +242 -0
- data/lib/BioDSL/command.rb +133 -0
- data/lib/BioDSL/commands/add_key.rb +110 -0
- data/lib/BioDSL/commands/align_seq_mothur.rb +194 -0
- data/lib/BioDSL/commands/analyze_residue_distribution.rb +222 -0
- data/lib/BioDSL/commands/assemble_pairs.rb +336 -0
- data/lib/BioDSL/commands/assemble_seq_idba.rb +230 -0
- data/lib/BioDSL/commands/assemble_seq_ray.rb +345 -0
- data/lib/BioDSL/commands/assemble_seq_spades.rb +252 -0
- data/lib/BioDSL/commands/classify_seq.rb +217 -0
- data/lib/BioDSL/commands/classify_seq_mothur.rb +226 -0
- data/lib/BioDSL/commands/clip_primer.rb +318 -0
- data/lib/BioDSL/commands/cluster_otus.rb +181 -0
- data/lib/BioDSL/commands/collapse_otus.rb +170 -0
- data/lib/BioDSL/commands/collect_otus.rb +150 -0
- data/lib/BioDSL/commands/complement_seq.rb +117 -0
- data/lib/BioDSL/commands/count.rb +135 -0
- data/lib/BioDSL/commands/count_values.rb +149 -0
- data/lib/BioDSL/commands/degap_seq.rb +253 -0
- data/lib/BioDSL/commands/dereplicate_seq.rb +168 -0
- data/lib/BioDSL/commands/dump.rb +157 -0
- data/lib/BioDSL/commands/filter_rrna.rb +239 -0
- data/lib/BioDSL/commands/genecall.rb +237 -0
- data/lib/BioDSL/commands/grab.rb +535 -0
- data/lib/BioDSL/commands/index_taxonomy.rb +226 -0
- data/lib/BioDSL/commands/mask_seq.rb +175 -0
- data/lib/BioDSL/commands/mean_scores.rb +168 -0
- data/lib/BioDSL/commands/merge_pair_seq.rb +175 -0
- data/lib/BioDSL/commands/merge_table.rb +225 -0
- data/lib/BioDSL/commands/merge_values.rb +113 -0
- data/lib/BioDSL/commands/plot_heatmap.rb +233 -0
- data/lib/BioDSL/commands/plot_histogram.rb +306 -0
- data/lib/BioDSL/commands/plot_matches.rb +282 -0
- data/lib/BioDSL/commands/plot_residue_distribution.rb +278 -0
- data/lib/BioDSL/commands/plot_scores.rb +285 -0
- data/lib/BioDSL/commands/random.rb +153 -0
- data/lib/BioDSL/commands/read_fasta.rb +222 -0
- data/lib/BioDSL/commands/read_fastq.rb +414 -0
- data/lib/BioDSL/commands/read_table.rb +329 -0
- data/lib/BioDSL/commands/reverse_seq.rb +113 -0
- data/lib/BioDSL/commands/slice_align.rb +400 -0
- data/lib/BioDSL/commands/slice_seq.rb +151 -0
- data/lib/BioDSL/commands/sort.rb +223 -0
- data/lib/BioDSL/commands/split_pair_seq.rb +220 -0
- data/lib/BioDSL/commands/split_values.rb +165 -0
- data/lib/BioDSL/commands/trim_primer.rb +314 -0
- data/lib/BioDSL/commands/trim_seq.rb +192 -0
- data/lib/BioDSL/commands/uchime_ref.rb +170 -0
- data/lib/BioDSL/commands/uclust.rb +286 -0
- data/lib/BioDSL/commands/unique_values.rb +145 -0
- data/lib/BioDSL/commands/usearch_global.rb +171 -0
- data/lib/BioDSL/commands/usearch_local.rb +171 -0
- data/lib/BioDSL/commands/write_fasta.rb +207 -0
- data/lib/BioDSL/commands/write_fastq.rb +191 -0
- data/lib/BioDSL/commands/write_table.rb +419 -0
- data/lib/BioDSL/commands/write_tree.rb +167 -0
- data/lib/BioDSL/commands.rb +31 -0
- data/lib/BioDSL/config.rb +55 -0
- data/lib/BioDSL/csv.rb +307 -0
- data/lib/BioDSL/debug.rb +42 -0
- data/lib/BioDSL/fasta.rb +133 -0
- data/lib/BioDSL/fastq.rb +77 -0
- data/lib/BioDSL/filesys.rb +137 -0
- data/lib/BioDSL/fork.rb +145 -0
- data/lib/BioDSL/hamming.rb +128 -0
- data/lib/BioDSL/helpers/aux_helper.rb +44 -0
- data/lib/BioDSL/helpers/email_helper.rb +66 -0
- data/lib/BioDSL/helpers/history_helper.rb +40 -0
- data/lib/BioDSL/helpers/log_helper.rb +55 -0
- data/lib/BioDSL/helpers/options_helper.rb +405 -0
- data/lib/BioDSL/helpers/status_helper.rb +132 -0
- data/lib/BioDSL/helpers.rb +35 -0
- data/lib/BioDSL/html_report.rb +200 -0
- data/lib/BioDSL/math.rb +55 -0
- data/lib/BioDSL/mummer.rb +216 -0
- data/lib/BioDSL/pipeline.rb +354 -0
- data/lib/BioDSL/seq/ambiguity.rb +66 -0
- data/lib/BioDSL/seq/assemble.rb +240 -0
- data/lib/BioDSL/seq/backtrack.rb +252 -0
- data/lib/BioDSL/seq/digest.rb +99 -0
- data/lib/BioDSL/seq/dynamic.rb +263 -0
- data/lib/BioDSL/seq/homopolymer.rb +59 -0
- data/lib/BioDSL/seq/kmer.rb +293 -0
- data/lib/BioDSL/seq/levenshtein.rb +113 -0
- data/lib/BioDSL/seq/translate.rb +109 -0
- data/lib/BioDSL/seq/trim.rb +188 -0
- data/lib/BioDSL/seq.rb +742 -0
- data/lib/BioDSL/serializer.rb +98 -0
- data/lib/BioDSL/stream.rb +113 -0
- data/lib/BioDSL/taxonomy.rb +691 -0
- data/lib/BioDSL/test.rb +42 -0
- data/lib/BioDSL/tmp_dir.rb +68 -0
- data/lib/BioDSL/usearch.rb +301 -0
- data/lib/BioDSL/verbose.rb +42 -0
- data/lib/BioDSL/version.rb +31 -0
- data/lib/BioDSL.rb +81 -0
- data/test/BioDSL/commands/test_add_key.rb +105 -0
- data/test/BioDSL/commands/test_align_seq_mothur.rb +99 -0
- data/test/BioDSL/commands/test_analyze_residue_distribution.rb +134 -0
- data/test/BioDSL/commands/test_assemble_pairs.rb +459 -0
- data/test/BioDSL/commands/test_assemble_seq_idba.rb +50 -0
- data/test/BioDSL/commands/test_assemble_seq_ray.rb +51 -0
- data/test/BioDSL/commands/test_assemble_seq_spades.rb +50 -0
- data/test/BioDSL/commands/test_classify_seq.rb +50 -0
- data/test/BioDSL/commands/test_classify_seq_mothur.rb +59 -0
- data/test/BioDSL/commands/test_clip_primer.rb +377 -0
- data/test/BioDSL/commands/test_cluster_otus.rb +128 -0
- data/test/BioDSL/commands/test_collapse_otus.rb +81 -0
- data/test/BioDSL/commands/test_collect_otus.rb +82 -0
- data/test/BioDSL/commands/test_complement_seq.rb +78 -0
- data/test/BioDSL/commands/test_count.rb +103 -0
- data/test/BioDSL/commands/test_count_values.rb +85 -0
- data/test/BioDSL/commands/test_degap_seq.rb +96 -0
- data/test/BioDSL/commands/test_dereplicate_seq.rb +92 -0
- data/test/BioDSL/commands/test_dump.rb +109 -0
- data/test/BioDSL/commands/test_filter_rrna.rb +128 -0
- data/test/BioDSL/commands/test_genecall.rb +50 -0
- data/test/BioDSL/commands/test_grab.rb +398 -0
- data/test/BioDSL/commands/test_index_taxonomy.rb +62 -0
- data/test/BioDSL/commands/test_mask_seq.rb +98 -0
- data/test/BioDSL/commands/test_mean_scores.rb +111 -0
- data/test/BioDSL/commands/test_merge_pair_seq.rb +115 -0
- data/test/BioDSL/commands/test_merge_table.rb +131 -0
- data/test/BioDSL/commands/test_merge_values.rb +83 -0
- data/test/BioDSL/commands/test_plot_heatmap.rb +185 -0
- data/test/BioDSL/commands/test_plot_histogram.rb +194 -0
- data/test/BioDSL/commands/test_plot_matches.rb +157 -0
- data/test/BioDSL/commands/test_plot_residue_distribution.rb +309 -0
- data/test/BioDSL/commands/test_plot_scores.rb +308 -0
- data/test/BioDSL/commands/test_random.rb +88 -0
- data/test/BioDSL/commands/test_read_fasta.rb +229 -0
- data/test/BioDSL/commands/test_read_fastq.rb +552 -0
- data/test/BioDSL/commands/test_read_table.rb +327 -0
- data/test/BioDSL/commands/test_reverse_seq.rb +79 -0
- data/test/BioDSL/commands/test_slice_align.rb +218 -0
- data/test/BioDSL/commands/test_slice_seq.rb +131 -0
- data/test/BioDSL/commands/test_sort.rb +128 -0
- data/test/BioDSL/commands/test_split_pair_seq.rb +164 -0
- data/test/BioDSL/commands/test_split_values.rb +95 -0
- data/test/BioDSL/commands/test_trim_primer.rb +329 -0
- data/test/BioDSL/commands/test_trim_seq.rb +150 -0
- data/test/BioDSL/commands/test_uchime_ref.rb +113 -0
- data/test/BioDSL/commands/test_uclust.rb +139 -0
- data/test/BioDSL/commands/test_unique_values.rb +98 -0
- data/test/BioDSL/commands/test_usearch_global.rb +123 -0
- data/test/BioDSL/commands/test_usearch_local.rb +125 -0
- data/test/BioDSL/commands/test_write_fasta.rb +159 -0
- data/test/BioDSL/commands/test_write_fastq.rb +166 -0
- data/test/BioDSL/commands/test_write_table.rb +411 -0
- data/test/BioDSL/commands/test_write_tree.rb +122 -0
- data/test/BioDSL/helpers/test_options_helper.rb +272 -0
- data/test/BioDSL/seq/test_assemble.rb +98 -0
- data/test/BioDSL/seq/test_backtrack.rb +176 -0
- data/test/BioDSL/seq/test_digest.rb +71 -0
- data/test/BioDSL/seq/test_dynamic.rb +133 -0
- data/test/BioDSL/seq/test_homopolymer.rb +58 -0
- data/test/BioDSL/seq/test_kmer.rb +134 -0
- data/test/BioDSL/seq/test_translate.rb +75 -0
- data/test/BioDSL/seq/test_trim.rb +101 -0
- data/test/BioDSL/test_cary.rb +176 -0
- data/test/BioDSL/test_command.rb +45 -0
- data/test/BioDSL/test_csv.rb +514 -0
- data/test/BioDSL/test_debug.rb +42 -0
- data/test/BioDSL/test_fasta.rb +154 -0
- data/test/BioDSL/test_fastq.rb +46 -0
- data/test/BioDSL/test_filesys.rb +145 -0
- data/test/BioDSL/test_fork.rb +85 -0
- data/test/BioDSL/test_math.rb +41 -0
- data/test/BioDSL/test_mummer.rb +79 -0
- data/test/BioDSL/test_pipeline.rb +187 -0
- data/test/BioDSL/test_seq.rb +790 -0
- data/test/BioDSL/test_serializer.rb +72 -0
- data/test/BioDSL/test_stream.rb +55 -0
- data/test/BioDSL/test_taxonomy.rb +336 -0
- data/test/BioDSL/test_test.rb +42 -0
- data/test/BioDSL/test_tmp_dir.rb +58 -0
- data/test/BioDSL/test_usearch.rb +33 -0
- data/test/BioDSL/test_verbose.rb +42 -0
- data/test/helper.rb +82 -0
- data/www/command.html.haml +14 -0
- data/www/css.html.haml +55 -0
- data/www/input_files.html.haml +3 -0
- data/www/layout.html.haml +12 -0
- data/www/output_files.html.haml +3 -0
- data/www/overview.html.haml +15 -0
- data/www/pipeline.html.haml +4 -0
- data/www/png.html.haml +2 -0
- data/www/status.html.haml +9 -0
- data/www/time.html.haml +11 -0
- metadata +503 -0
@@ -0,0 +1,239 @@
|
|
1
|
+
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< #
|
2
|
+
# #
|
3
|
+
# Copyright (C) 2007-2015 Martin Asser Hansen (mail@maasha.dk). #
|
4
|
+
# #
|
5
|
+
# This program is free software; you can redistribute it and/or #
|
6
|
+
# modify it under the terms of the GNU General Public License #
|
7
|
+
# as published by the Free Software Foundation; either version 2 #
|
8
|
+
# of the License, or (at your option) any later version. #
|
9
|
+
# #
|
10
|
+
# This program is distributed in the hope that it will be useful, #
|
11
|
+
# but WITHOUT ANY WARRANTY; without even the implied warranty of #
|
12
|
+
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the #
|
13
|
+
# GNU General Public License for more details. #
|
14
|
+
# #
|
15
|
+
# You should have received a copy of the GNU General Public License #
|
16
|
+
# along with this program; if not, write to the Free Software #
|
17
|
+
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, #
|
18
|
+
# USA. #
|
19
|
+
# #
|
20
|
+
# http://www.gnu.org/copyleft/gpl.html #
|
21
|
+
# #
|
22
|
+
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< #
|
23
|
+
# #
|
24
|
+
# This software is part of the BioDSL framework (www.BioDSL.org). #
|
25
|
+
# #
|
26
|
+
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< #
|
27
|
+
|
28
|
+
module BioDSL
|
29
|
+
# == Filter rRNA sequences from the stream.
|
30
|
+
#
|
31
|
+
# Description
|
32
|
+
#
|
33
|
+
# +filter_rrna+ utilizes +sortmerna+ to identify and filter ribosomal RNA
|
34
|
+
# sequences from the stream. The +sortmerna+ and +indexdb_rna+ executables
|
35
|
+
# must be installed for +filter_rrna+ to work.
|
36
|
+
#
|
37
|
+
# Indexed reference files are produced using +indexdb_rna+.
|
38
|
+
#
|
39
|
+
# For more about the sortmerna look here:
|
40
|
+
#
|
41
|
+
# http://bioinfo.lifl.fr/RNA/sortmerna/
|
42
|
+
#
|
43
|
+
# == Usage
|
44
|
+
# filter_rrna(ref_fasta: <file(s)>, ref_index: <file(s)>)
|
45
|
+
#
|
46
|
+
# === Options
|
47
|
+
# * ref_fasta <file(s)> - One or more reference FASTA files.
|
48
|
+
# * ref_index <file(s)> - One or more index reference files.
|
49
|
+
#
|
50
|
+
# == Examples
|
51
|
+
#
|
52
|
+
# To filter all reads matching the SILVA archaea 23S rRNA do:
|
53
|
+
#
|
54
|
+
# BP.new.
|
55
|
+
# read_fastq(input: "reads.fq").
|
56
|
+
# filter_rrna(ref_fasta: ["silva-arc-23s-id98.fasta"],
|
57
|
+
# ref_index: ["silva-arc-23s-id98.fasta.idx*"]).
|
58
|
+
# write_fastq(output: "clean.fq").
|
59
|
+
# run
|
60
|
+
#
|
61
|
+
# rubocop:disable ClassLength
|
62
|
+
class FilterRrna
|
63
|
+
require 'English'
|
64
|
+
require 'set'
|
65
|
+
require 'BioDSL/helpers/aux_helper'
|
66
|
+
|
67
|
+
include AuxHelper
|
68
|
+
|
69
|
+
STATS = %i(records_in records_out sequences_in sequences_out residues_in
|
70
|
+
residues_out)
|
71
|
+
|
72
|
+
# Constructor the FilterRrna class.
|
73
|
+
#
|
74
|
+
# @param options [Hash] Options hash.
|
75
|
+
# @option options [String,Array] Path(s) to reference FASTA files.
|
76
|
+
# @option options [String,Array] Path(s) to reference index files.
|
77
|
+
#
|
78
|
+
# @return [FilterRrnas] Class instance of FilterRrnas.
|
79
|
+
def initialize(options)
|
80
|
+
@options = options
|
81
|
+
@filter = Set.new
|
82
|
+
|
83
|
+
aux_exist('sortmerna')
|
84
|
+
check_options
|
85
|
+
end
|
86
|
+
|
87
|
+
# Return the command lambda for filter_rrnas.
|
88
|
+
#
|
89
|
+
# @return [Proc] Command lambda.
|
90
|
+
def lmb
|
91
|
+
lambda do |input, output, status|
|
92
|
+
status_init(status, STATS)
|
93
|
+
|
94
|
+
TmpDir.create('tmp', 'seq', 'out') do |tmp_file, seq_file, out_file|
|
95
|
+
ref_files = process_ref_files
|
96
|
+
process_input(input, tmp_file, seq_file)
|
97
|
+
execute_sortmerna(ref_files, seq_file, out_file)
|
98
|
+
parse_sortme_output(out_file)
|
99
|
+
process_output(output, tmp_file)
|
100
|
+
end
|
101
|
+
end
|
102
|
+
end
|
103
|
+
|
104
|
+
private
|
105
|
+
|
106
|
+
# Check options.
|
107
|
+
def check_options
|
108
|
+
options_allowed(@options, :ref_fasta, :ref_index)
|
109
|
+
options_files_exist(@options, :ref_fasta, :ref_index)
|
110
|
+
end
|
111
|
+
|
112
|
+
# Given reference index and fasta files in the options hash, process these
|
113
|
+
# into a string of the format read by 'sortmerna': fasta1,id1:fasta2,id2:...
|
114
|
+
#
|
115
|
+
# @return [String] Reference file string for sortmerna.
|
116
|
+
def process_ref_files
|
117
|
+
ref_index = @options[:ref_index]
|
118
|
+
ref_fasta = @options[:ref_fasta]
|
119
|
+
|
120
|
+
if ref_index.is_a? Array
|
121
|
+
ref_index.map { |f| f.sub!(/\*$/, '') }
|
122
|
+
else
|
123
|
+
ref_index.sub!(/\*$/, '')
|
124
|
+
end
|
125
|
+
|
126
|
+
ref_fasta = [ref_fasta.split(',')] if ref_fasta.is_a? String
|
127
|
+
ref_index = [ref_index.split(',')] if ref_index.is_a? String
|
128
|
+
|
129
|
+
ref_fasta.zip(ref_index).map { |m| m.join(',') }.join(':')
|
130
|
+
end
|
131
|
+
|
132
|
+
# Execute 'sortmerna'.
|
133
|
+
#
|
134
|
+
# @param ref_files [String] Reference file string for sortmerna.
|
135
|
+
# @param seq_file [String] Path to intput file with reads.
|
136
|
+
# @param out_file [String] Path to output file.
|
137
|
+
#
|
138
|
+
# @raise if execution of 'sortmerna' fails.
|
139
|
+
def execute_sortmerna(ref_files, seq_file, out_file)
|
140
|
+
cmd = ['sortmerna']
|
141
|
+
cmd << "--ref #{ref_files}"
|
142
|
+
cmd << "--reads #{seq_file}"
|
143
|
+
cmd << "--aligned #{out_file}"
|
144
|
+
cmd << '--fastx'
|
145
|
+
cmd << '-v' if BioDSL.verbose
|
146
|
+
|
147
|
+
cmd_line = cmd.join(' ')
|
148
|
+
|
149
|
+
$stderr.puts "Running command: #{cmd_line}" if BioDSL.verbose
|
150
|
+
|
151
|
+
system(cmd_line)
|
152
|
+
|
153
|
+
fail "command failed: #{cmd_line}" unless $CHILD_STATUS.success?
|
154
|
+
end
|
155
|
+
|
156
|
+
# Parse the 'sortmerna' output file and add all sequence name indices to the
|
157
|
+
# filter set.
|
158
|
+
#
|
159
|
+
# @param out_file [String] Path to output file.
|
160
|
+
def parse_sortme_output(out_file)
|
161
|
+
BioDSL::Fasta.open("#{out_file}.fasta", 'r') do |ios|
|
162
|
+
ios.each do |entry|
|
163
|
+
@filter << entry.seq_name.to_i
|
164
|
+
end
|
165
|
+
end
|
166
|
+
end
|
167
|
+
|
168
|
+
# Process input stream and serialize all records and write a temporary FASTA
|
169
|
+
# file.
|
170
|
+
#
|
171
|
+
# @param input [Enumerator] Input stream.
|
172
|
+
# @param tmp_file [String] Path to tmp file for serialized records.
|
173
|
+
# @param seq_file [String] Path to tmp FASTA sequence file.
|
174
|
+
def process_input(input, tmp_file, seq_file)
|
175
|
+
BioDSL::Fasta.open(seq_file, 'w') do |seq_io|
|
176
|
+
File.open(tmp_file, 'wb') do |tmp_ios|
|
177
|
+
BioDSL::Serializer.new(tmp_ios) do |s|
|
178
|
+
input.each_with_index do |record, i|
|
179
|
+
@status[:records_in] += 1
|
180
|
+
|
181
|
+
s << record
|
182
|
+
# FIXME: need << method
|
183
|
+
seq_io.puts record2entry(record, i).to_fasta if record.key? :SEQ
|
184
|
+
end
|
185
|
+
end
|
186
|
+
end
|
187
|
+
end
|
188
|
+
end
|
189
|
+
|
190
|
+
# Given a BioDSL record and an index create a new sequence entry object
|
191
|
+
# that is returned using the index as sequence name.
|
192
|
+
#
|
193
|
+
# @param record [Hash] BioDSL record
|
194
|
+
# @param i [Integer] Index.
|
195
|
+
#
|
196
|
+
# @return [BioDSL::Seq] Sequence entry.
|
197
|
+
def record2entry(record, i)
|
198
|
+
entry = BioDSL::Seq.new(seq_name: i, seq: record[:SEQ])
|
199
|
+
@status[:sequences_in] += 1
|
200
|
+
@status[:residues_in] += entry.length
|
201
|
+
entry
|
202
|
+
end
|
203
|
+
|
204
|
+
# Process the serialized data and output all records, that does not match
|
205
|
+
# the filter, to the output stream.
|
206
|
+
#
|
207
|
+
# @param output [Enumerator::Yielder] Output stream.
|
208
|
+
# @param tmp_file [String] Path to tmp file with serialized records.
|
209
|
+
def process_output(output, tmp_file)
|
210
|
+
File.open(tmp_file, 'rb') do |ios|
|
211
|
+
BioDSL::Serializer.new(ios) do |s|
|
212
|
+
s.each_with_index do |record, i|
|
213
|
+
output_record(output, record, i)
|
214
|
+
end
|
215
|
+
end
|
216
|
+
end
|
217
|
+
end
|
218
|
+
|
219
|
+
# Output a record to the output stream unless it contains sequence
|
220
|
+
# information that should be filtered.
|
221
|
+
#
|
222
|
+
# @param output [Enumerator::Yielder] Output stream.
|
223
|
+
# @param record [Hash] BioDSL record.
|
224
|
+
# @param i [Integer] Index.
|
225
|
+
def output_record(output, record, i)
|
226
|
+
if record.key? :SEQ
|
227
|
+
unless @filter.include? i
|
228
|
+
output << record
|
229
|
+
@status[:records_out] += 1
|
230
|
+
@status[:sequences_out] += 1
|
231
|
+
@status[:residues_out] += record[:SEQ].length
|
232
|
+
end
|
233
|
+
else
|
234
|
+
output << record
|
235
|
+
@status[:records_out] += 1
|
236
|
+
end
|
237
|
+
end
|
238
|
+
end
|
239
|
+
end
|
@@ -0,0 +1,237 @@
|
|
1
|
+
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< #
|
2
|
+
# #
|
3
|
+
# Copyright (C) 2007-2015 Martin Asser Hansen (mail@maasha.dk). #
|
4
|
+
# #
|
5
|
+
# This program is free software; you can redistribute it and/or #
|
6
|
+
# modify it under the terms of the GNU General Public License #
|
7
|
+
# as published by the Free Software Foundation; either version 2 #
|
8
|
+
# of the License, or (at your option) any later version. #
|
9
|
+
# #
|
10
|
+
# This program is distributed in the hope that it will be useful, #
|
11
|
+
# but WITHOUT ANY WARRANTY; without even the implied warranty of #
|
12
|
+
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the #
|
13
|
+
# GNU General Public License for more details. #
|
14
|
+
# #
|
15
|
+
# You should have received a copy of the GNU General Public License #
|
16
|
+
# along with this program; if not, write to the Free Software #
|
17
|
+
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, #
|
18
|
+
# USA. #
|
19
|
+
# #
|
20
|
+
# http://www.gnu.org/copyleft/gpl.html #
|
21
|
+
# #
|
22
|
+
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< #
|
23
|
+
# #
|
24
|
+
# This software is part of the BioDSL framework (www.BioDSL.org). #
|
25
|
+
# #
|
26
|
+
# >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>><<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< #
|
27
|
+
|
28
|
+
module BioDSL
|
29
|
+
# == Genecall sequences in the stream.
|
30
|
+
#
|
31
|
+
# +Genecall+ predict genes in prokaryotic single genomes or metagenomes using
|
32
|
+
# Prodigal 2.6 which must be installed:
|
33
|
+
#
|
34
|
+
# http://prodigal.ornl.gov/
|
35
|
+
#
|
36
|
+
# The records produced are of the type:
|
37
|
+
#
|
38
|
+
# {:RECORD_TYPE=>"gene",
|
39
|
+
# :S_BEG=>2, :S_END=>109,
|
40
|
+
# :S_LEN=>108,
|
41
|
+
# :STRAND=>"-",
|
42
|
+
# :SEQ_NAME=>"contig1",
|
43
|
+
# :SEQ=>"MGKVIGIDLGTTNSCVAVMDGKTAKVIENAEGMRTT",
|
44
|
+
# :SEQ_LEN=>36}
|
45
|
+
#
|
46
|
+
# == Usage
|
47
|
+
#
|
48
|
+
# genecall([type: <string>[, procedure: <string>[, closed_ends: <bool>
|
49
|
+
# [, masked: <bool>]]]])
|
50
|
+
#
|
51
|
+
# === Options
|
52
|
+
#
|
53
|
+
# * type: <string> - Output dna or protein sequence (default: dna).
|
54
|
+
# * procedure: <string> - Single or meta (default: single).
|
55
|
+
# * closed_ends: <bool> - Don't allow truncated gene at ends.
|
56
|
+
# * masked: <bool> - Ignore stretch of Ns.
|
57
|
+
#
|
58
|
+
# == Examples
|
59
|
+
#
|
60
|
+
# To genecall a genome do:
|
61
|
+
#
|
62
|
+
# BP.new.
|
63
|
+
# read_fasta(input: "contigs.fna").
|
64
|
+
# genecall.
|
65
|
+
# grab(select: "genecall", key: :type, exact: true).
|
66
|
+
# write_fasta(output: "genes.fna").
|
67
|
+
# run
|
68
|
+
#
|
69
|
+
# To add genecall data to the sequence name use +merge_values+:
|
70
|
+
#
|
71
|
+
# BP.new.
|
72
|
+
# read_fasta(input: "contigs.fna").
|
73
|
+
# genecall(type: "protein").
|
74
|
+
# grab(select: "genecall", key: :type, exact: true).
|
75
|
+
# merge_values(keys: [:SEQ_NAME, :S_BEG, :S_END, :S_LEN, :STRAND]).
|
76
|
+
# write_fasta(output: "genes.faa").
|
77
|
+
# run
|
78
|
+
class Genecall
|
79
|
+
require 'English'
|
80
|
+
require 'BioDSL/helpers/aux_helper'
|
81
|
+
|
82
|
+
include AuxHelper
|
83
|
+
|
84
|
+
STATS = %i(records_in records_out sequences_in sequences_out residues_in
|
85
|
+
residues_out)
|
86
|
+
|
87
|
+
# Constructor for the Genecall class.
|
88
|
+
#
|
89
|
+
# @param [Hash] options Options hash.
|
90
|
+
# @option options [Symbol] :type of output.
|
91
|
+
# @option options [Symbol] :procedure used for genecalling.
|
92
|
+
# @option options [Boolean] :closed_ends disallow truncated genes at ends.
|
93
|
+
# @option options [Boolean] :masked ignore stretch of Ns.
|
94
|
+
#
|
95
|
+
# @return [Genecall] Returns an instance of the class.
|
96
|
+
def initialize(options)
|
97
|
+
@options = options
|
98
|
+
@names = {}
|
99
|
+
|
100
|
+
aux_exist('prodigal')
|
101
|
+
defaults
|
102
|
+
check_options
|
103
|
+
|
104
|
+
@type = @options[:type].to_sym
|
105
|
+
end
|
106
|
+
|
107
|
+
# Return a lambda for the genecall command.
|
108
|
+
#
|
109
|
+
# @return [Proc] Returns the command lambda.
|
110
|
+
def lmb
|
111
|
+
lambda do |input, output, status|
|
112
|
+
status_init(status, STATS)
|
113
|
+
|
114
|
+
TmpDir.create('i.fa', 'o.fna', 'o.faa') do |tmp_in, tmp_fna, tmp_faa|
|
115
|
+
process_input(input, output, tmp_in)
|
116
|
+
run_prodigal(tmp_in, tmp_fna, tmp_faa)
|
117
|
+
process_output(output, tmp_fna, tmp_faa)
|
118
|
+
end
|
119
|
+
end
|
120
|
+
end
|
121
|
+
|
122
|
+
private
|
123
|
+
|
124
|
+
# Run Prodigal on the input file.
|
125
|
+
#
|
126
|
+
# @param tmp_in [String] Path to input FASTA file.
|
127
|
+
# @param tmp_fna [String] Path to output FASTA DNA file.
|
128
|
+
# @param tmp_faa [String] Path to output FASTA Protein file.
|
129
|
+
def run_prodigal(tmp_in, tmp_fna, tmp_faa)
|
130
|
+
cmd = []
|
131
|
+
cmd << 'prodigal'
|
132
|
+
cmd << '-f gff'
|
133
|
+
cmd << '-c' if @options[:closed_ends]
|
134
|
+
cmd << '-m' if @options[:masked]
|
135
|
+
cmd << "-p #{@options[:procedure]}"
|
136
|
+
cmd << "-i #{tmp_in}"
|
137
|
+
cmd << "-d #{tmp_fna}"
|
138
|
+
cmd << "-a #{tmp_faa}"
|
139
|
+
cmd << '-q' unless BioDSL.verbose
|
140
|
+
cmd << '> /dev/null 2>&1' unless BioDSL.verbose
|
141
|
+
|
142
|
+
cmd_line = cmd.join(' ')
|
143
|
+
|
144
|
+
$stderr.puts "Running: #{cmd_line}" if BioDSL.verbose
|
145
|
+
system(cmd_line)
|
146
|
+
|
147
|
+
fail cmd_line unless $CHILD_STATUS.success?
|
148
|
+
end
|
149
|
+
|
150
|
+
# Check the options.
|
151
|
+
def check_options
|
152
|
+
options_allowed(@options, :type, :procedure, :closed_ends, :masked)
|
153
|
+
options_allowed_values(@options, type: [:dna, :protein, 'dna',
|
154
|
+
'protein'])
|
155
|
+
options_allowed_values(@options, procedure: ['single', 'meta', :single,
|
156
|
+
:meta])
|
157
|
+
options_allowed_values(@options, closed_ends: [nil, true, false])
|
158
|
+
options_allowed_values(@options, masked: [nil, true, false])
|
159
|
+
end
|
160
|
+
|
161
|
+
# Set the default option values.
|
162
|
+
def defaults
|
163
|
+
@options[:type] ||= :dna
|
164
|
+
@options[:procedure] ||= :single
|
165
|
+
end
|
166
|
+
|
167
|
+
# Read all records from input and emit non-sequence records to the output
|
168
|
+
# stream. Sequence records are saved to a temporary file.
|
169
|
+
#
|
170
|
+
# @param input [Enumerator] input stream.
|
171
|
+
# @param output [Enumerator::Yielder] Output stream.
|
172
|
+
# @param fa_in [String] Path to temporary FASTA file.
|
173
|
+
def process_input(input, output, fa_in)
|
174
|
+
BioDSL::Fasta.open(fa_in, 'w') do |fasta_io|
|
175
|
+
input.each_with_index do |record, i|
|
176
|
+
@status[:records_in] += 1
|
177
|
+
|
178
|
+
if record.key? :SEQ
|
179
|
+
entry = BioDSL::Seq.new(seq_name: i, seq: record[:SEQ])
|
180
|
+
@names[i] = record[:SEQ_NAME] || i
|
181
|
+
|
182
|
+
@status[:sequences_in] += 1
|
183
|
+
@status[:sequences_out] += 1
|
184
|
+
@status[:residues_in] += entry.length
|
185
|
+
@status[:residues_out] += entry.length
|
186
|
+
|
187
|
+
fasta_io.puts entry.to_fasta
|
188
|
+
end
|
189
|
+
|
190
|
+
@status[:records_out] += 1
|
191
|
+
output << record
|
192
|
+
end
|
193
|
+
end
|
194
|
+
end
|
195
|
+
|
196
|
+
# Read the output from file and emit to the output stream.
|
197
|
+
#
|
198
|
+
# @param output [Enumerator::Yielder] Output stream.
|
199
|
+
# @param tmp_fna [String] Path to output FASTA DNA file.
|
200
|
+
# @param tmp_faa [String] Path to output FASTA Protein file.
|
201
|
+
def process_output(output, tmp_fna, tmp_faa)
|
202
|
+
file = (@type == :dna) ? tmp_fna : tmp_faa
|
203
|
+
|
204
|
+
BioDSL::Fasta.open(file, 'r') do |ios|
|
205
|
+
ios.each do |entry|
|
206
|
+
output << parse_entry(entry)
|
207
|
+
|
208
|
+
@status[:records_out] += 1
|
209
|
+
@status[:sequences_out] += 1
|
210
|
+
@status[:residues_out] += entry.length
|
211
|
+
end
|
212
|
+
end
|
213
|
+
end
|
214
|
+
|
215
|
+
# Parse Prodigal genecall data from sequence name.
|
216
|
+
#
|
217
|
+
# @param entry [BioDSL::Seq] Sequence object.
|
218
|
+
#
|
219
|
+
# @return [Hash] BioPiece record.
|
220
|
+
def parse_entry(entry)
|
221
|
+
record = {}
|
222
|
+
fields = entry.seq_name.split(' # ')
|
223
|
+
|
224
|
+
record[:RECORD_TYPE] = 'genecall'
|
225
|
+
record[:S_BEG] = fields[1].to_i - 1
|
226
|
+
record[:S_END] = fields[2].to_i - 1
|
227
|
+
record[:S_LEN] = record[:S_END] - record[:S_BEG] + 1
|
228
|
+
record[:STRAND] = fields[3] == '1' ? '+' : '-'
|
229
|
+
record[:SEQ_NAME] = @names[fields[0].split('_').first.to_i]
|
230
|
+
record[:SEQ] = entry.seq
|
231
|
+
record[:SEQ_LEN] = entry.length
|
232
|
+
record[:SEQ_TYPE] = @type.to_s
|
233
|
+
|
234
|
+
record
|
235
|
+
end
|
236
|
+
end
|
237
|
+
end
|