bio-polyploid-tools 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/Gemfile +16 -0
- data/Gemfile.lock +67 -0
- data/README +21 -0
- data/Rakefile +61 -0
- data/VERSION +1 -0
- data/bin/bfr.rb +133 -0
- data/bin/count_variations.rb +36 -0
- data/bin/filter_blat_by_target_coverage.rb +15 -0
- data/bin/find_best_blat_hit.rb +32 -0
- data/bin/hexaploid_primers.rb +168 -0
- data/bin/homokaryot_primers.rb +155 -0
- data/bin/map_markers_to_contigs.rb +66 -0
- data/bin/markers_in_region.rb +42 -0
- data/bin/polymarker.rb +219 -0
- data/bin/snps_between_bams.rb +106 -0
- data/bio-polyploid-tools.gemspec +139 -0
- data/conf/defaults.rb +1 -0
- data/conf/primer3_config/dangle.dh +128 -0
- data/conf/primer3_config/dangle.ds +128 -0
- data/conf/primer3_config/interpretations/dangle_i.dh +131 -0
- data/conf/primer3_config/interpretations/dangle_i.ds +131 -0
- data/conf/primer3_config/interpretations/loops_i.dh +34 -0
- data/conf/primer3_config/interpretations/loops_i.ds +31 -0
- data/conf/primer3_config/interpretations/stack_i.dh +257 -0
- data/conf/primer3_config/interpretations/stack_i.ds +256 -0
- data/conf/primer3_config/interpretations/stackmm_i_mm.dh +257 -0
- data/conf/primer3_config/interpretations/stackmm_i_mm.ds +256 -0
- data/conf/primer3_config/interpretations/tetraloop_i.dh +79 -0
- data/conf/primer3_config/interpretations/tetraloop_i.ds +81 -0
- data/conf/primer3_config/interpretations/triloop_i.dh +21 -0
- data/conf/primer3_config/interpretations/triloop_i.ds +18 -0
- data/conf/primer3_config/interpretations/tstack2_i.dh +256 -0
- data/conf/primer3_config/interpretations/tstack2_i.ds +256 -0
- data/conf/primer3_config/interpretations/tstack_i.dh +256 -0
- data/conf/primer3_config/interpretations/tstack_i.ds +256 -0
- data/conf/primer3_config/interpretations/tstack_tm_inf_i.dh +256 -0
- data/conf/primer3_config/interpretations/tstack_tm_inf_i.ds +256 -0
- data/conf/primer3_config/loops.dh +30 -0
- data/conf/primer3_config/loops.ds +30 -0
- data/conf/primer3_config/stack.dh +256 -0
- data/conf/primer3_config/stack.ds +256 -0
- data/conf/primer3_config/stackmm.dh +256 -0
- data/conf/primer3_config/stackmm.ds +256 -0
- data/conf/primer3_config/tetraloop.dh +77 -0
- data/conf/primer3_config/tetraloop.ds +77 -0
- data/conf/primer3_config/triloop.dh +16 -0
- data/conf/primer3_config/triloop.ds +16 -0
- data/conf/primer3_config/tstack.dh +256 -0
- data/conf/primer3_config/tstack2.dh +256 -0
- data/conf/primer3_config/tstack2.ds +256 -0
- data/conf/primer3_config/tstack_tm_inf.ds +256 -0
- data/lib/bio/BFRTools.rb +698 -0
- data/lib/bio/BIOExtensions.rb +186 -0
- data/lib/bio/PolyploidTools/ChromosomeArm.rb +52 -0
- data/lib/bio/PolyploidTools/ExonContainer.rb +194 -0
- data/lib/bio/PolyploidTools/Marker.rb +175 -0
- data/lib/bio/PolyploidTools/PrimerRegion.rb +22 -0
- data/lib/bio/PolyploidTools/SNP.rb +681 -0
- data/lib/bio/PolyploidTools/SNPSequence.rb +56 -0
- data/lib/bio/SAMToolsExtensions.rb +284 -0
- data/lib/bio/db/exonerate.rb +272 -0
- data/lib/bio/db/fastadb.rb +164 -0
- data/lib/bio/db/primer3.rb +673 -0
- data/lib/bioruby-polyploid-tools.rb +25 -0
- data/test/data/BS00068396_51.fa +2 -0
- data/test/data/BS00068396_51_contigs.aln +1412 -0
- data/test/data/BS00068396_51_contigs.dnd +7 -0
- data/test/data/BS00068396_51_contigs.fa +8 -0
- data/test/data/BS00068396_51_exonerate.tab +6 -0
- data/test/data/BS00068396_51_genes.txt +14 -0
- data/test/data/LIB1716.bam +0 -0
- data/test/data/LIB1716.bam.bai +0 -0
- data/test/data/LIB1719.bam +0 -0
- data/test/data/LIB1719.bam.bai +0 -0
- data/test/data/LIB1721.bam +0 -0
- data/test/data/LIB1721.bam.bai +0 -0
- data/test/data/LIB1722.bam +0 -0
- data/test/data/LIB1722.bam.bai +0 -0
- data/test/data/S22380157.fa +16 -0
- data/test/data/S22380157.fa.fai +1 -0
- data/test/data/Test3Aspecific.csv +1 -0
- data/test/data/Test3Aspecific_contigs.fa +6 -0
- data/test/data/patological_cases5D.csv +1 -0
- data/test/data/short_primer_design_test.csv +10 -0
- data/test/data/test_primer3_error.csv +4 -0
- data/test/data/test_primer3_error_contigs.fa +10 -0
- data/test/test_bfr.rb +51 -0
- data/test/test_exon_container.rb +17 -0
- data/test/test_exonearate.rb +53 -0
- data/test/test_snp_parsing.rb +40 -0
- metadata +201 -0
data/lib/bio/BFRTools.rb
ADDED
@@ -0,0 +1,698 @@
|
|
1
|
+
require 'rubygems'
|
2
|
+
#require 'extensions/all'
|
3
|
+
#require 'bio-samtools'
|
4
|
+
#require 'bio/db/pileup'
|
5
|
+
#require 'bio/db/vcf'
|
6
|
+
require 'pathname'
|
7
|
+
#require_relative 'BIOExtensions.rb'
|
8
|
+
require_relative 'db/fastadb.rb'
|
9
|
+
|
10
|
+
require 'bio'
|
11
|
+
require "set"
|
12
|
+
require 'systemu'
|
13
|
+
require 'json'
|
14
|
+
#require 'strmask'
|
15
|
+
|
16
|
+
=begin
|
17
|
+
|
18
|
+
Extends the methods to be able to calculate the BFR and a consensus from the pileup
|
19
|
+
|
20
|
+
=end
|
21
|
+
|
22
|
+
class Bio::DB::Pileup
|
23
|
+
|
24
|
+
#attr_accessor :minumum_ratio_for_iup_consensus
|
25
|
+
#@minumum_ratio_for_iup_consensus = 0.20
|
26
|
+
|
27
|
+
#Returns a hash with the count of bases
|
28
|
+
|
29
|
+
def bases
|
30
|
+
return @bases if @bases
|
31
|
+
@bases = self.non_refs
|
32
|
+
#puts self.ref_count
|
33
|
+
@bases[self.ref_base.upcase.to_sym] = self.ref_count
|
34
|
+
@bases
|
35
|
+
end
|
36
|
+
|
37
|
+
def base_coverage
|
38
|
+
total = 0
|
39
|
+
@bases.each do |k,v|
|
40
|
+
total += v
|
41
|
+
end
|
42
|
+
total
|
43
|
+
end
|
44
|
+
|
45
|
+
def base_ratios
|
46
|
+
return @base_ratios if @base_ratios
|
47
|
+
bases = self.bases
|
48
|
+
@base_ratios = Hash.new
|
49
|
+
bases.each do |k,v|
|
50
|
+
@base_ratios[k] = v.to_f/self.base_coverage.to_f
|
51
|
+
end
|
52
|
+
@base_ratios
|
53
|
+
end
|
54
|
+
|
55
|
+
# returns the consensus (most frequent) base from the pileup, if there are equally represented bases returns a string of all equally represented bases in alphabetical order
|
56
|
+
def consensus_iuap(minumum_ratio_for_iup_consensus)
|
57
|
+
minumum_ratio_for_iup_consensus
|
58
|
+
if @consensus_iuap.nil?
|
59
|
+
@consensus_iuap = self.ref_base.downcase
|
60
|
+
bases = self.bases
|
61
|
+
tmp = String.new
|
62
|
+
bases.each do |k,v|
|
63
|
+
tmp << k[0].to_s if v/self.coverage > minumum_ratio_for_iup_consensus
|
64
|
+
end
|
65
|
+
if tmp.length > 0
|
66
|
+
@consensus_iuap = Bio::NucleicAcid.to_IUAPC(tmp)
|
67
|
+
end
|
68
|
+
end
|
69
|
+
@consensus_iuap
|
70
|
+
end
|
71
|
+
end
|
72
|
+
|
73
|
+
class Bio::DB::Fasta::Region
|
74
|
+
attr_accessor :pileup, :average_coverage, :snps, :reference, :base_ratios, :consensus, :coverages, :bases
|
75
|
+
|
76
|
+
#TODO: Debug, as it hasnt been tested in the actual code.
|
77
|
+
def base_ratios_for_base(base)
|
78
|
+
@all_ratios = Hash.new unless @all_ratios
|
79
|
+
unless @all_ratios[base]
|
80
|
+
ratios = Array.new
|
81
|
+
for i in (0..region.size-1)
|
82
|
+
ratios << @base_ratios[i][base]
|
83
|
+
end
|
84
|
+
@all_ratios[base] = ratios
|
85
|
+
end
|
86
|
+
@all_ratios[base]
|
87
|
+
end
|
88
|
+
|
89
|
+
end
|
90
|
+
|
91
|
+
class Bio::DB::Sam::SAMException < RuntimeError
|
92
|
+
|
93
|
+
end
|
94
|
+
|
95
|
+
class Bio::DB::Sam
|
96
|
+
|
97
|
+
|
98
|
+
attr_accessor :minumum_ratio_for_iup_consensus
|
99
|
+
attr_reader :cached_regions
|
100
|
+
#attr_accessor :pileup_cache
|
101
|
+
@minumum_ratio_for_iup_consensus = 0.20
|
102
|
+
|
103
|
+
|
104
|
+
#Same as mpilup, but it caches the pileup, so if you want several operations on the same set of regions
|
105
|
+
#the pile for different operations, it won't execute the mpilup command several times
|
106
|
+
#Whenever you finish using a region, call mpileup_clear_cache to free the cache
|
107
|
+
#The argument Region is required, as it will be the key for the underlying hash.
|
108
|
+
#We asume that the options are constant. If they are not, the cache mechanism may not be consistent.
|
109
|
+
#
|
110
|
+
#TODO: It may be good to load partially the pileup
|
111
|
+
def mpileup_cached (opts={})
|
112
|
+
raise SAMException.new(), "A region must be provided" unless opts[:r] or opts[:region]
|
113
|
+
@pileup_cache = Hash.new unless @pileup_cache
|
114
|
+
@cached_regions = Hash.new unless @cached_regions
|
115
|
+
|
116
|
+
region = opts[:r] ? opts[:r] : opts[:region]
|
117
|
+
opts[:r] = "'#{region.to_s}'"
|
118
|
+
opts[:region] = "'#{region.to_s}'"
|
119
|
+
opts[:A] = true
|
120
|
+
#reg = region.class == Bio::DB::Fasta::Region ? region : Bio::DB::Fasta::Region.parse_region(region.to_s)
|
121
|
+
|
122
|
+
unless @cached_regions[region.to_s]
|
123
|
+
@cached_regions[region.to_s] = Bio::DB::Fasta::Region.parse_region(region.to_s)
|
124
|
+
tmp = Array.new
|
125
|
+
@cached_regions[region.to_s].pileup = tmp
|
126
|
+
#puts "Loading #{region.to_s}"
|
127
|
+
mpileup(opts) do | pile |
|
128
|
+
# puts pile
|
129
|
+
tmp << pile
|
130
|
+
yield pile
|
131
|
+
end
|
132
|
+
else
|
133
|
+
# puts "Loaded, reruning #{region.to_s}"
|
134
|
+
@cached_regions.pileup[region.to_s] .each do | pile |
|
135
|
+
yield pile
|
136
|
+
end
|
137
|
+
end
|
138
|
+
end
|
139
|
+
|
140
|
+
#Clears the pileup cache. If a region is passed as argument, just the specified region is removed
|
141
|
+
#If no region is passed, the hash is emptied
|
142
|
+
def mpileup_clear_cache (region)
|
143
|
+
return unless @cached_regions
|
144
|
+
if region
|
145
|
+
@cached_regions[region.to_s] = nil
|
146
|
+
else
|
147
|
+
@cached_regions.clear
|
148
|
+
end
|
149
|
+
end
|
150
|
+
|
151
|
+
#Gets the coverage of a region from a pileup.
|
152
|
+
def average_coverage_from_pileup(opts={})
|
153
|
+
opts[:region] = opts[:region].to_s if opts[:region] .class == Bio::DB::Fasta::Region
|
154
|
+
region = opts[:region]
|
155
|
+
calculate_stats_from_pile(opts) if @cached_regions == nil or @cached_regions[region] == nil
|
156
|
+
@cached_regions[region].average_coverage
|
157
|
+
end
|
158
|
+
|
159
|
+
#
|
160
|
+
def coverages_from_pileup(opts={})
|
161
|
+
opts[:region] = opts[:region].to_s if opts[:region] .class == Bio::DB::Fasta::Region
|
162
|
+
region = opts[:region]
|
163
|
+
calculate_stats_from_pile(opts) if @cached_regions == nil or @cached_regions[region] == nil
|
164
|
+
@cached_regions[region].coverages
|
165
|
+
end
|
166
|
+
|
167
|
+
def consensus_with_ambiguities(opts={})
|
168
|
+
opts[:region] = opts[:region].to_s if opts[:region] .class == Bio::DB::Fasta::Region
|
169
|
+
region = opts[:region]
|
170
|
+
# p "consensus with ambiguities for: " << opts[:region]
|
171
|
+
calculate_stats_from_pile(opts) if @cached_regions == nil or @cached_regions[region] == nil
|
172
|
+
@cached_regions[region].consensus
|
173
|
+
end
|
174
|
+
|
175
|
+
def calculate_stats_from_pile(opts={})
|
176
|
+
min_cov = opts[:min_cov] ? opts[:min_cov] : 20
|
177
|
+
|
178
|
+
|
179
|
+
opts[:region] = Bio::DB::Fasta::Region.parse_region( opts[:region] .to_s) unless opts[:region].class == Bio::DB::Fasta::Region
|
180
|
+
region = opts[:region]
|
181
|
+
reference = self.fetch_reference(region.entry, region.start, region.end).downcase
|
182
|
+
# p "calculationg from pile..." << region.to_s
|
183
|
+
base_ratios = Array.new(region.size, BASE_COUNT_ZERO)
|
184
|
+
bases = Array.new(region.size, BASE_COUNT_ZERO)
|
185
|
+
coverages = Array.new(region.size, 0)
|
186
|
+
total_cov = 0
|
187
|
+
|
188
|
+
self.mpileup_cached(:region=>"#{region.to_s}") do | pile |
|
189
|
+
#puts pile
|
190
|
+
#puts pile.coverage
|
191
|
+
if pile.coverage > min_cov
|
192
|
+
base_ratios[pile.pos - region.start ] = pile.base_ratios
|
193
|
+
reference[pile.pos - region.start ] = pile.consensus_iuap(0.20)
|
194
|
+
coverages[pile.pos - region.start ] = pile.coverage.to_i
|
195
|
+
bases[pile.pos - region.start ] = pile.bases
|
196
|
+
end
|
197
|
+
total_cov += pile.coverage
|
198
|
+
end
|
199
|
+
|
200
|
+
region = @cached_regions[region.to_s]
|
201
|
+
region.coverages = coverages
|
202
|
+
region.base_ratios = base_ratios
|
203
|
+
region.consensus = reference
|
204
|
+
|
205
|
+
region.average_coverage = total_cov.to_f/region.size.to_f
|
206
|
+
region.bases = bases
|
207
|
+
region
|
208
|
+
end
|
209
|
+
|
210
|
+
|
211
|
+
|
212
|
+
BASE_COUNT_ZERO = {:A => 0, :C => 0, :G => 0, :T => 0}
|
213
|
+
|
214
|
+
#Gets an array with the proportions of the bases in the region. If there is no coverage, a
|
215
|
+
def base_ratios_in_region(opts={})
|
216
|
+
opts[:region] = opts[:region].to_s if opts[:region] .class == Bio::DB::Fasta::Region
|
217
|
+
region = opts[:region]
|
218
|
+
calculate_stats_from_pile(opts) if @cached_regions == nil or @cached_regions[region] == nil
|
219
|
+
@cached_regions[region].base_ratios
|
220
|
+
end
|
221
|
+
|
222
|
+
#Gets an array with the bsaes count in the region. If there is no coverage, a
|
223
|
+
def bases_in_region(opts={})
|
224
|
+
opts[:region] = opts[:region].to_s if opts[:region] .class == Bio::DB::Fasta::Region
|
225
|
+
region = opts[:region]
|
226
|
+
calculate_stats_from_pile(opts) if @cached_regions == nil or @cached_regions[region] == nil
|
227
|
+
@cached_regions[region].bases
|
228
|
+
end
|
229
|
+
|
230
|
+
|
231
|
+
|
232
|
+
def extract_reads(opts={})
|
233
|
+
opts[:region] = Bio::DB::Fasta::Region.parse_region( opts[:region] .to_s) unless opts[:region].class == Bio::DB::Fasta::Region
|
234
|
+
fastq_filename = opts[:fastq]
|
235
|
+
fastq_file = opts[:fastq_file]
|
236
|
+
|
237
|
+
out = $stdout
|
238
|
+
|
239
|
+
print_fastq = Proc.new do |alignment|
|
240
|
+
out.puts "@#{alignment.qname}"
|
241
|
+
out.puts "#{alignment.seq}"
|
242
|
+
out.puts "+#{alignment.qname}"
|
243
|
+
out.puts "#{alignment.qual}"
|
244
|
+
end
|
245
|
+
|
246
|
+
fetch_with_function(chromosome, qstart, qstart+len, print_fastq)
|
247
|
+
|
248
|
+
|
249
|
+
end
|
250
|
+
|
251
|
+
|
252
|
+
|
253
|
+
end
|
254
|
+
|
255
|
+
module Bio::BFRTools
|
256
|
+
|
257
|
+
|
258
|
+
|
259
|
+
class BFRToolsException < StandardError; end
|
260
|
+
|
261
|
+
|
262
|
+
class Container
|
263
|
+
|
264
|
+
attr_reader :putative_snps, :processed_regions, :total_length, :parental_1_sam, :parental_2_sam, :bulk_1_sam, :bulk_2_sam
|
265
|
+
attr_reader :parental_1_name, :parental_2_name, :bulk_1_name, :bulk_2_name, :reference_db
|
266
|
+
|
267
|
+
BASES = [:A, :C, :G, :T]
|
268
|
+
#Sets the reference file
|
269
|
+
def reference(path)
|
270
|
+
@reference_db = Bio::DB::Fasta::FastaFile.new(path)
|
271
|
+
@reference_path = path
|
272
|
+
end
|
273
|
+
|
274
|
+
def reference_sequence(region)
|
275
|
+
@reference_db.fetch_sequence(region)
|
276
|
+
end
|
277
|
+
|
278
|
+
#Sets the sorted BAM file of the first parental
|
279
|
+
#It accepts the following arguments
|
280
|
+
#:name=>A name for thie parental 1 (optional). If not provided,
|
281
|
+
#:path=>
|
282
|
+
def parental_1(opts)
|
283
|
+
raise BFRToolsException.new("Missing path for parental 1") if opts[:path] == nil
|
284
|
+
path = Pathname.new(opts[:path])
|
285
|
+
raise BFRToolsException.new("Unable to open #{path}") unless path.readable? or path.directory?
|
286
|
+
|
287
|
+
@parental_1_name = opts[:name] ? opts[:name] : path.basename(".bam").to_s
|
288
|
+
@parental_1_sam = Bio::DB::Sam.new({:fasta=>@reference_path, :bam=>path.realpath.to_s})
|
289
|
+
@parental_1_path = path
|
290
|
+
|
291
|
+
end
|
292
|
+
|
293
|
+
#Sets the sorted BAM file of the second parental
|
294
|
+
def parental_2(opts)
|
295
|
+
raise BFRToolsException.new("Missing path for parental 2") if opts[:path] == nil
|
296
|
+
path = Pathname.new(opts[:path])
|
297
|
+
raise BFRToolsException.new("Unable to open #{path}") unless path.readable? or path.directory?
|
298
|
+
|
299
|
+
@parental_2_name = @name = opts[:name] ? opts[:name] : path.basename(".bam").to_s
|
300
|
+
@parental_2_sam = Bio::DB::Sam.new({:fasta=>@reference_path, :bam=>path.realpath.to_s})
|
301
|
+
@parental_2_path = path
|
302
|
+
end
|
303
|
+
|
304
|
+
#Sets the sorted BAM file of the first bulk
|
305
|
+
def bulk_1(opts)
|
306
|
+
raise BFRToolsException.new("Missing path for bulk 1") if opts[:path] == nil
|
307
|
+
path = Pathname.new(opts[:path])
|
308
|
+
raise BFRToolsException.new("Unable to open #{path}") unless path.readable? or path.directory?
|
309
|
+
|
310
|
+
@bulk_1_name = opts[:name] ? opts[:name] : path.basename(".bam").to_s
|
311
|
+
@bulk_1_sam = Bio::DB::Sam.new({:fasta=>@reference_path, :bam=>path.realpath.to_s})
|
312
|
+
@bulk_1_path = path
|
313
|
+
end
|
314
|
+
|
315
|
+
#Sets the sorted BAM file of the second bulk
|
316
|
+
def bulk_2(opts)
|
317
|
+
raise BFRToolsException.new("Missing path for bulk 2") if opts[:path] == nil
|
318
|
+
path = Pathname.new(opts[:path])
|
319
|
+
raise BFRToolsException.new("Unable to open #{path}") unless path.readable? or path.directory?
|
320
|
+
|
321
|
+
@bulk_2_name = opts[:name] ? opts[:name] : path.basename(".bam").to_s
|
322
|
+
@bulk_2_sam = Bio::DB::Sam.new({:fasta=>@reference_path, :bam=>path.realpath.to_s})
|
323
|
+
@bulk_2_path = path
|
324
|
+
end
|
325
|
+
|
326
|
+
|
327
|
+
def self.snps_between(seq1, seq2)
|
328
|
+
snps=0
|
329
|
+
for i in (0..seq1.size)
|
330
|
+
snps += 1 if seq1[i] != seq2[i]
|
331
|
+
end
|
332
|
+
snps
|
333
|
+
end
|
334
|
+
end
|
335
|
+
|
336
|
+
class BFRLine
|
337
|
+
attr_reader :original_base, :variation_base, :position, :bulk_1_ratio, :bulk_2_ratio, :bfr
|
338
|
+
|
339
|
+
|
340
|
+
end
|
341
|
+
|
342
|
+
class BFRRegion < Bio::DB::Fasta::Region
|
343
|
+
BASES = [:A, :C, :G, :T]
|
344
|
+
attr_reader :parental_1_sequence, :parental_2_sequence, :bulk_1_sequence, :bulk_2_sequence, :snp_count
|
345
|
+
attr_reader :ratios_bulk_1, :ratios_bulk_2, :avg_cov_bulk_1, :avg_cov_bulk_2, :coverages_1, :coverages_2, :bases_bulk_1, :bases_bulk_2
|
346
|
+
|
347
|
+
def initialize(opts)
|
348
|
+
opts = { :min_cov=>20, :max_snp_1kbp => 5 }.merge!(opts)
|
349
|
+
reg = Bio::DB::Fasta::Region.parse_region(opts[:region])
|
350
|
+
self.entry = reg.entry
|
351
|
+
self.start = reg.start
|
352
|
+
self.end = reg.end
|
353
|
+
|
354
|
+
@container = opts[:container]
|
355
|
+
|
356
|
+
parental_1_sam = @container.parental_1_sam
|
357
|
+
parental_2_sam = @container.parental_2_sam
|
358
|
+
bulk_1_sam = @container.bulk_1_sam
|
359
|
+
bulk_2_sam = @container.bulk_2_sam
|
360
|
+
|
361
|
+
@parental_1_sequence = parental_1_sam.consensus_with_ambiguities(opts)
|
362
|
+
@parental_2_sequence = parental_2_sam.consensus_with_ambiguities(opts)
|
363
|
+
|
364
|
+
@bulk_1_sequence = bulk_1_sam.consensus_with_ambiguities(opts)
|
365
|
+
@bulk_2_sequence = bulk_2_sam.consensus_with_ambiguities(opts)
|
366
|
+
|
367
|
+
@snp_count = Container.snps_between( @parental_1_sequence , @parental_2_sequence )
|
368
|
+
|
369
|
+
@ratios_bulk_1 = bulk_1_sam.base_ratios_in_region(opts)
|
370
|
+
@ratios_bulk_2 = bulk_2_sam.base_ratios_in_region(opts)
|
371
|
+
|
372
|
+
@bases_bulk_1 = bulk_1_sam.bases_in_region(opts)
|
373
|
+
@bases_bulk_2 = bulk_2_sam.bases_in_region(opts)
|
374
|
+
|
375
|
+
@avg_cov_bulk_1 = bulk_1_sam.average_coverage_from_pileup(opts)
|
376
|
+
@avg_cov_bulk_2 = bulk_2_sam.average_coverage_from_pileup(opts)
|
377
|
+
|
378
|
+
@coverages_1 = bulk_1_sam.coverages_from_pileup(opts)
|
379
|
+
@coverages_2 = bulk_2_sam.coverages_from_pileup(opts)
|
380
|
+
|
381
|
+
end
|
382
|
+
|
383
|
+
def get_bfr_lines(opts = {})
|
384
|
+
|
385
|
+
opts = { :min_cov=>20, :max_snp_1kbp => 5 }.merge!(opts)
|
386
|
+
p opts.inspect
|
387
|
+
region = self
|
388
|
+
line = String.new
|
389
|
+
info = Array.new
|
390
|
+
|
391
|
+
for i in (0..region.size-1)
|
392
|
+
|
393
|
+
if region.coverages_1[i] > opts[:min_cov] and region.coverages_2[i] > opts[:min_cov]
|
394
|
+
BASES.each do |base|
|
395
|
+
|
396
|
+
info.clear
|
397
|
+
if Bio::NucleicAcid.is_valid( region.parental_1_sequence[i], base.to_s ) and
|
398
|
+
not Bio::NucleicAcid.is_valid( region.parental_2_sequence[i], base.to_s )
|
399
|
+
info << :first
|
400
|
+
end
|
401
|
+
|
402
|
+
if Bio::NucleicAcid.is_valid( region.parental_2_sequence[i], base.to_s ) and
|
403
|
+
not Bio::NucleicAcid.is_valid( region.parental_1_sequence[i], base.to_s )
|
404
|
+
info << :second
|
405
|
+
end
|
406
|
+
|
407
|
+
|
408
|
+
for informative in info
|
409
|
+
l = region.get_bfr_line(i, base, informative)
|
410
|
+
puts l << "\n"
|
411
|
+
line << l << "\n"
|
412
|
+
|
413
|
+
# output.print line , "\n"
|
414
|
+
end
|
415
|
+
end
|
416
|
+
end
|
417
|
+
end
|
418
|
+
line
|
419
|
+
end
|
420
|
+
|
421
|
+
|
422
|
+
def snp_1kbp
|
423
|
+
@snp_count.to_f * 1000 / self.size.to_f
|
424
|
+
end
|
425
|
+
|
426
|
+
def bfrs
|
427
|
+
return @BFRs if @BFRs
|
428
|
+
@BFRs = Hash.new
|
429
|
+
|
430
|
+
[:first, :second].each do | reference |
|
431
|
+
@BFRs[reference] = Hash.new
|
432
|
+
BASES.each do |base|
|
433
|
+
@BFRs[reference][base] = Array.new
|
434
|
+
end
|
435
|
+
end
|
436
|
+
|
437
|
+
|
438
|
+
for i in (0..self.size-1)
|
439
|
+
ratios_1 = @ratios_bulk_1[i]
|
440
|
+
ratios_2 = @ratios_bulk_2[i]
|
441
|
+
BASES.each do |base|
|
442
|
+
|
443
|
+
if ratios_1[base] == 0 and ratios_2[base] == 0
|
444
|
+
bfr1 = 0
|
445
|
+
bfr2 = 0
|
446
|
+
elsif ratios_1[base] == 0
|
447
|
+
bfr1 = 0
|
448
|
+
bfr2 = Float::INFINITY
|
449
|
+
elsif ratios_2[base] == 0
|
450
|
+
bfr1 = Float::INFINITY
|
451
|
+
bfr2 = 0
|
452
|
+
#bfr = Float::INFINITY
|
453
|
+
else
|
454
|
+
bfr1 = ratios_1[base] / ratios_2[base]
|
455
|
+
bfr2 = ratios_2[base] / ratios_1[base]
|
456
|
+
end
|
457
|
+
@BFRs[:first][base] << bfr1
|
458
|
+
@BFRs[:second][base] << bfr2
|
459
|
+
end
|
460
|
+
end
|
461
|
+
@BFRs
|
462
|
+
end
|
463
|
+
|
464
|
+
def get_bfr_line(position, base, reference)
|
465
|
+
if(reference == :first)
|
466
|
+
informative = @container.parental_1_name
|
467
|
+
ref_base = @parental_2_sequence[position]
|
468
|
+
elsif(reference == :second )
|
469
|
+
informative = @container.parental_2_name
|
470
|
+
ref_base = @parental_1_sequence[position]
|
471
|
+
else
|
472
|
+
raise BFRToolsException.new ("The reference for the line should be :first or :second, but was " + reference.to_s )
|
473
|
+
end
|
474
|
+
|
475
|
+
relative_position = self.start + position + 1
|
476
|
+
|
477
|
+
bfr = bfrs[reference][base][position]
|
478
|
+
cov_1 = @coverages_1[position]
|
479
|
+
cov_2 = @coverages_2[position]
|
480
|
+
ratios_1 = @ratios_bulk_1[position][base]
|
481
|
+
ratios_2 = @ratios_bulk_2[position][base]
|
482
|
+
base_1_count = @bases_bulk_1[position][base.to_sym]
|
483
|
+
base_2_count = @bases_bulk_2[position][base.to_sym]
|
484
|
+
#puts "bases_1 #{@bases_bulk_1[position].to_s}"
|
485
|
+
#puts "bases_2 #{@bases_bulk_2[position].to_s}"
|
486
|
+
line = String.new
|
487
|
+
line << @container.parental_1_name << "\t" << @container.parental_2_name << "\t" << @container.bulk_1_name << "\t" << @container.bulk_2_name << "\t" << self.entry << "\t"
|
488
|
+
line << ref_base << "\t" << relative_position.to_s
|
489
|
+
line << "\t" << base.to_s << "\t"
|
490
|
+
line << bfr.round(2).to_s << "\t"
|
491
|
+
line << cov_1.to_s << "\t" << cov_2.to_s << "\t"
|
492
|
+
line << informative
|
493
|
+
line << "\t" << ratios_1.round(2).to_s << "\t" << ratios_2.round(2).to_s
|
494
|
+
line << "\t" << base_1_count.to_s << "\t" << base_2_count.to_s
|
495
|
+
line
|
496
|
+
end
|
497
|
+
|
498
|
+
def to_multi_fasta
|
499
|
+
fasta_string = String.new
|
500
|
+
fasta_string << ">"<< self.to_s << ":" << @container.parental_1_name << "\n" << @parental_1_sequence << "\n"
|
501
|
+
fasta_string << ">"<< self.to_s << ":" << @container.parental_2_name << "\n" << @parental_2_sequence << "\n"
|
502
|
+
fasta_string << ">"<< self.to_s << ":" << @container.bulk_1_name << "\n" << @bulk_1_sequence << "\n"
|
503
|
+
fasta_string << ">"<< self.to_s << ":" << @container.bulk_2_name << "\n" << @bulk_2_sequence << "\n"
|
504
|
+
fasta_string
|
505
|
+
end
|
506
|
+
|
507
|
+
def to_json (opts)
|
508
|
+
# puts JSON.dump self
|
509
|
+
# JSON.dump self
|
510
|
+
#{}"{\"firstName\": \"John\"}"
|
511
|
+
out = String.new
|
512
|
+
out << "{"
|
513
|
+
out << "\"Parental_1\" : \"" << @container.parental_1_name << "\"\n"
|
514
|
+
out << "\"Parental 2\" : \"" << @container.parental_2_name << "\"\n"
|
515
|
+
out << "\"Bulk 1\" : \"" << @container.bulk_1_name << "\"\n"
|
516
|
+
out << "\"Bulk 2\" : \"" << @container.bulk_2_name << "\"\n"
|
517
|
+
out << "\"Positions\" : " << (1..self.size).to_a.to_json << "\n" #TODO: Make this for any subsection, so we can subquery in case we are working on something bigger
|
518
|
+
out << "\"Parental_1_consensus\":" << @parental_1_sequence .split(//).to_json << "\n"
|
519
|
+
out << "\"Parental_2_consensus\":" << @parental_2_sequence .split(//).to_json << "\n"
|
520
|
+
out << "\"Bulk_1_consensus\":" << @bulk_1_sequence .split(//).to_json << "\n"
|
521
|
+
out << "\"Bulk_1_coverage\":" << @coverages_1.to_json << "\n"
|
522
|
+
# puts BASES
|
523
|
+
|
524
|
+
BASES.each do |base|
|
525
|
+
out << "\"Bases_Bulk_1" << base.to_s << "\":" << base_count_for_base(base, @bases_bulk_1).join(",") << "\n"
|
526
|
+
out << "\"Ratios_Bulk_1" << base.to_s << "\":" << base_ratios_for_base(base, @ratios_bulk_1).join(",") << "\n"
|
527
|
+
end
|
528
|
+
out << "\"Bulk_2_consensus\":" << @bulk_2_sequence .split(//).join(",") << "\n"
|
529
|
+
out << "\"Bulk_2_coverage\":" << @coverages_2.join(",") << "\n"
|
530
|
+
|
531
|
+
BASES.each do |base|
|
532
|
+
out << "\"Bases_Bulk_2"<< base.to_s << "\":" << base_count_for_base(base, @bases_bulk_2).join(",") << "\n"
|
533
|
+
out << "\"Ratios_Bulk_2" << base.to_s << "\":" << base_ratios_for_base(base, @ratios_bulk_2).join(",") << "\n"
|
534
|
+
end
|
535
|
+
BASES.each do |base|
|
536
|
+
out << "\"BFR" << base.to_s << "\":" << bfrs[:first][base].join(",") << "\n"
|
537
|
+
end
|
538
|
+
# << "\t" << @container.bulk_2_name << "\t" << self.entry << "\t"
|
539
|
+
out << "}"
|
540
|
+
out
|
541
|
+
|
542
|
+
end
|
543
|
+
|
544
|
+
def to_csv
|
545
|
+
out = String.new
|
546
|
+
out << "Parental 1," << @container.parental_1_name << "\n"
|
547
|
+
out << "Parental 2," << @container.parental_2_name << "\n"
|
548
|
+
out << "Bulk 1, " << @container.bulk_1_name << "\n"
|
549
|
+
out << "Bulk 2," << @container.bulk_2_name << "\n"
|
550
|
+
out << "Positions," << (1..self.size).to_a.join(",") << "\n"
|
551
|
+
out << "Parental 1 consensus," << @parental_1_sequence .split(//).join(",") << "\n"
|
552
|
+
out << "Parental 2 consensus," << @parental_2_sequence .split(//).join(",") << "\n"
|
553
|
+
out << "Bulk 1 consensus," << @bulk_1_sequence .split(//).join(",") << "\n"
|
554
|
+
out << "Bulk 1 coverage," << @coverages_1.join(",") << "\n"
|
555
|
+
# puts BASES
|
556
|
+
BASES.each do |base|
|
557
|
+
out << "Bases Bulk 1"<< base.to_s << "," << base_count_for_base(base, @bases_bulk_1).join(",") << "\n"
|
558
|
+
out << "Ratios Bulk 1 " << base.to_s << "," << base_ratios_for_base(base, @ratios_bulk_1).join(",") << "\n"
|
559
|
+
end
|
560
|
+
out << "Bulk 2 consensus," << @bulk_2_sequence .split(//).join(",") << "\n"
|
561
|
+
out << "Bulk 2 coverage," << @coverages_2.join(",") << "\n"
|
562
|
+
|
563
|
+
BASES.each do |base|
|
564
|
+
out << "Bases Bulk 2 "<< base.to_s << "," << base_count_for_base(base, @bases_bulk_2).join(",") << "\n"
|
565
|
+
out << "Ratios Bulk 2 " << base.to_s << "," << base_ratios_for_base(base, @ratios_bulk_2).join(",") << "\n"
|
566
|
+
end
|
567
|
+
BASES.each do |base|
|
568
|
+
out << "BFRs" << base.to_s << "," << bfrs[:first][base].join(",") << "\n"
|
569
|
+
end
|
570
|
+
# << "\t" << @container.bulk_2_name << "\t" << self.entry << "\t"
|
571
|
+
out
|
572
|
+
end
|
573
|
+
|
574
|
+
def base_ratios_for_base(base, ratios_matrix)
|
575
|
+
ratios = Array.new
|
576
|
+
for i in (0..ratios_matrix.size-1)
|
577
|
+
ratios << ratios_matrix[i][base]
|
578
|
+
end
|
579
|
+
ratios
|
580
|
+
end
|
581
|
+
|
582
|
+
def base_count_for_base(base, base_matrix)
|
583
|
+
bases = Array.new
|
584
|
+
for i in (0..base_matrix.size-1)
|
585
|
+
bases << base_matrix[i][base]
|
586
|
+
end
|
587
|
+
bases
|
588
|
+
end
|
589
|
+
|
590
|
+
end
|
591
|
+
|
592
|
+
|
593
|
+
class BFRContainer < Container
|
594
|
+
|
595
|
+
def init_counters
|
596
|
+
@putative_snps = 0
|
597
|
+
@proccesed_regions = 0
|
598
|
+
@not_enogh_coverage = 0
|
599
|
+
@total_avg_coverage_bulk_1 = 0.0
|
600
|
+
@total_avg_coverage_bulk_2 = 0.0
|
601
|
+
@total_snp_1kbp = 0.0
|
602
|
+
@no_snps = 0
|
603
|
+
@too_many_snps = 0
|
604
|
+
|
605
|
+
end
|
606
|
+
def print_header(opts={})
|
607
|
+
output = opts[:output_file_stats] ? opts[:output_file_stats] : $stderr
|
608
|
+
output.print "#bulk_1\tbulk_2\tProcessed_regions\tputative_snps\tno_snps\ttoo_many_snps\tno_enough_coverage\tavg_cov_bulk_1\tavg_cov_bulk_2\tavg_snp_1kbp\n"
|
609
|
+
end
|
610
|
+
|
611
|
+
def print_stats(opts={})
|
612
|
+
output = opts[:output_file_stats] ? opts[:output_file_stats] : $stderr
|
613
|
+
output.print @bulk_1_name, "\t", @bulk_2_name, "\t"
|
614
|
+
output.print @proccesed_regions, "\t", @putative_snps, "\t", @no_snps, "\t", @too_many_snps,"\t", @not_enogh_coverage, "\t"
|
615
|
+
output.print @total_avg_coverage_bulk_1/@proccesed_regions, "\t",@total_avg_coverage_bulk_2/@proccesed_regions, "\t"
|
616
|
+
output.print @total_snp_1kbp / @proccesed_regions,"\n"
|
617
|
+
end
|
618
|
+
|
619
|
+
def get_region(opts={})
|
620
|
+
opts[:container] = self
|
621
|
+
region = BFRRegion.new(opts)
|
622
|
+
end
|
623
|
+
|
624
|
+
def process_region(opts={})
|
625
|
+
opts = { :min_cov=>20, :max_snp_1kbp => 10 }.merge!(opts)
|
626
|
+
|
627
|
+
@proccesed_regions += 1
|
628
|
+
output = opts[:output_file] ? opts[:output_file] : $stdout
|
629
|
+
print_output = opts[:output_file] ? true : false
|
630
|
+
opts[:container] = self
|
631
|
+
|
632
|
+
region = BFRRegion.new(opts)
|
633
|
+
|
634
|
+
#puts region.to_multi_fasta
|
635
|
+
|
636
|
+
@total_snp_1kbp += region.snp_1kbp
|
637
|
+
# puts "SNPS: #{region.snp_1kbp}"
|
638
|
+
if region.snp_count == 0
|
639
|
+
@no_snps += 1
|
640
|
+
print_output = false
|
641
|
+
end
|
642
|
+
|
643
|
+
if region.snp_1kbp > opts[:max_snp_1kbp]
|
644
|
+
@too_many_snps += 1
|
645
|
+
print_output = false
|
646
|
+
end
|
647
|
+
|
648
|
+
|
649
|
+
|
650
|
+
@total_avg_coverage_bulk_2 += region.avg_cov_bulk_2
|
651
|
+
@total_avg_coverage_bulk_1 += region.avg_cov_bulk_1
|
652
|
+
|
653
|
+
if region.avg_cov_bulk_2 < opts[:min_cov] or region.avg_cov_bulk_1 < opts[:min_cov]
|
654
|
+
@not_enogh_coverage += 1
|
655
|
+
print_output = false
|
656
|
+
end
|
657
|
+
|
658
|
+
info = Array.new
|
659
|
+
|
660
|
+
if print_output
|
661
|
+
for i in (0..region.size-1)
|
662
|
+
if region.coverages_1[i] > opts[:min_cov] and region.coverages_2[i] > opts[:min_cov]
|
663
|
+
BASES.each do |base|
|
664
|
+
|
665
|
+
info.clear
|
666
|
+
if Bio::NucleicAcid.is_valid( region.parental_1_sequence[i], base.to_s ) and
|
667
|
+
not Bio::NucleicAcid.is_valid( region.parental_2_sequence[i], base.to_s )
|
668
|
+
info << :first
|
669
|
+
end
|
670
|
+
|
671
|
+
if Bio::NucleicAcid.is_valid( region.parental_2_sequence[i], base.to_s ) and
|
672
|
+
not Bio::NucleicAcid.is_valid( region.parental_1_sequence[i], base.to_s )
|
673
|
+
info << :second
|
674
|
+
end
|
675
|
+
|
676
|
+
|
677
|
+
for informative in info
|
678
|
+
line = region.get_bfr_line(i+1, base, informative)
|
679
|
+
output.print line , "\n"
|
680
|
+
end
|
681
|
+
end
|
682
|
+
end
|
683
|
+
end
|
684
|
+
end
|
685
|
+
|
686
|
+
|
687
|
+
@parental_1_sam.mpileup_clear_cache region
|
688
|
+
@parental_2_sam.mpileup_clear_cache region
|
689
|
+
@bulk_2_sam.mpileup_clear_cache region
|
690
|
+
@bulk_1_sam.mpileup_clear_cache region
|
691
|
+
return region
|
692
|
+
end
|
693
|
+
|
694
|
+
end
|
695
|
+
|
696
|
+
|
697
|
+
end
|
698
|
+
|