transrate 0.2.0 → 0.3.0

data/lib/transrate/contig_metrics.rb

@@ -1,5 +1,4 @@
 require 'set'
-require 'inline'
 
 module Transrate
 
@@ -41,7 +40,7 @@ module Transrate
       cpg_count = 0
       lc = 0
       k = 6
-      @assembly.assembly.each do |contig|
+      @assembly.assembly.each_value do |contig|
         total += contig.length
         a += contig.bases_a
         c += contig.bases_c

data/lib/transrate/read_metrics.rb

@@ -196,6 +196,8 @@ module Transrate
       CSV.open('supported_bridges.csv', 'w') do |f|
         @bridges.each_pair do |b, count|
           start, finish = b.to_s.split('<>')
+          @assembly[start].in_bridges += 1
+          @assembly[finish].in_bridges += 1
           if count > 1
             f << [start, finish, count]
             @supported_bridges += 1
@@ -215,15 +217,18 @@ module Transrate
       n, tot_length, tot_coverage = 0, 0, 0
       @assembly.each_with_coverage(bam) do |contig, coverage|
         next if contig.length < 200
-        zerocov, total = 0, 0
-        coverage.each { |e| total += e; zerocov += 1 if e < 1 }
+        contig.uncovered_bases, total = 0, 0
+        coverage.each do |e|
+          total += e
+          contig.uncovered_bases += 1 if e < 1
+        end
         tot_length += coverage.length
         tot_coverage += total
-        mean = total / coverage.length.to_f
-        @n_uncovered_bases += zerocov
-        @n_uncovered_base_contigs += 1 if zerocov > 0
-        @n_uncovered_contigs += 1 if mean < 1
-        @n_lowcovered_contigs += 1 if mean < 10
+        contig.mean_coverage = total / coverage.length.to_f
+        @n_uncovered_bases += contig.uncovered_bases
+        @n_uncovered_base_contigs += 1 if contig.uncovered_bases > 0
+        @n_uncovered_contigs += 1 if contig.mean_coverage < 1
+        @n_lowcovered_contigs += 1 if contig.mean_coverage < 10
       end
       @mean_coverage = (tot_coverage / tot_length.to_f).round(2)
       @p_uncovered_bases = @n_uncovered_bases / @assembly.n_bases.to_f
@@ -236,3 +241,4 @@ module Transrate
   end # ReadMetrics
 
 end # Transrate
+

data/lib/transrate/version.rb

@@ -9,7 +9,7 @@ module Transrate
   # Semantic Versioning 2.0 (http://semver.org/).
   module VERSION
     MAJOR = 0
-    MINOR = 2
+    MINOR = 3
     PATCH = 0
     BUILD = nil
 

data/test/helper.rb

@@ -11,37 +11,7 @@ require 'minitest/autorun'
 begin; require 'turn/autorun'; rescue LoadError; end
 require 'shoulda/context'
 require 'transrate'
+require 'transrate/transrate.so'
 
 Turn.config.format = :pretty
 Turn.config.trace = 5
-
-# fake CRBBlast class
-class CRBHelper
-
-  attr_accessor :target_is_prot, :hash
-  def initialize t
-    @target_is_prot = t
-  end
-
-  def reciprocals
-    return @hash
-  end
-
-end
-
-# rake Hit class
-class HitHelper
-
-  attr_accessor :query, :target, :qstart, :qend, :tstart, :tend, :qlen, :tlen
-  def initialize query, target, qstart, qend, tstart, tend, qlen, tlen
-    @query = query
-    @target = target
-    @qstart = qstart
-    @tstart = tstart
-    @tend = tend
-    @qend = qend
-    @qlen = qlen
-    @tlen = tlen
-  end
-
-end

data/test/test_bin.rb

@@ -0,0 +1,99 @@
+require 'helper'
+require 'csv'
+
+class TestTransrateBin < Test::Unit::TestCase
+
+  context "Transrate" do
+
+    setup do
+
+    end
+
+    teardown do
+      files = ["150uncovered.l.fq.150uncovered.r.fq.assembly.2.bai",
+      "150uncovered.l.fq.150uncovered.r.fq.assembly.2.bam",
+      "150uncovered.l.fq.150uncovered.r.fq.assembly.2.sam",
+      "150uncovered.l.fq.150uncovered.r.fq.assembly.2.sorted.bam",
+      "assembly.2.1.bt2", "assembly.2.2.bt2", "assembly.2.3.bt2",
+      "assembly.2.4.bt2", "assembly.2.fa.coverage",
+      "assembly.2_into_Os.protein.2.1.blast",
+      "assembly.2.nhr", "assembly.2.nin", "assembly.2.nsq",
+      "assembly.2.rev.1.bt2",  "assembly.2.rev.2.bt2",
+      "Os.protein.2_into_assembly.2.2.blast",
+      "Os.protein.2.phr",  "Os.protein.2.pin",  "Os.protein.2.psq",
+      "supported_bridges.csv",
+      "transrate_assemblies.csv",
+      "transrate_contigs.csv"]
+      files.each do |file|
+        File.delete(file) if File.exist?(file)
+      end
+    end
+
+    should "run help" do
+      c=Transrate::Cmd.new("bundle exec bin/transrate --help")
+      c.run
+      assert_equal 1751, c.stdout.length, "stdout"
+      assert_equal true, c.status.success?, "exit status"
+    end
+
+    should "fail on non existent assembly files" do
+      c=Transrate::Cmd.new("bundle exec bin/transrate --assembly foo.fasta")
+      c.run
+      assert_equal 163, c.stderr.length, "stderr"
+      assert_equal false, c.status.success?, "exit success"
+    end
+
+    should "fail on non existent reference files" do
+      c=Transrate::Cmd.new("bundle exec bin/transrate --reference foo.fasta")
+      c.run
+      assert_equal 104, c.stderr.length, "error"
+      assert_equal false, c.status.success?, "exit status"
+    end
+
+    should "run on test data" do
+      assembly = File.join(File.dirname(__FILE__), 'data', 'assembly.2.fa')
+      reference = File.join(File.dirname(__FILE__), 'data', 'Os.protein.2.fa')
+      left = File.join(File.dirname(__FILE__), 'data', '150uncovered.l.fq')
+      right = File.join(File.dirname(__FILE__), 'data', '150uncovered.r.fq')
+      cmd = "bundle exec bin/transrate --assembly #{assembly}"
+      cmd << " --reference #{reference}"
+      cmd << " --left #{left}"
+      cmd << " --right #{right}"
+      c = Transrate::Cmd.new("#{cmd}")
+      c.run
+      assert_equal true, c.status.success?, "exit status"
+      assert File.exist?("transrate_assemblies.csv"), "csv file doesn't exit"
+      assert File.exist?("transrate_contigs.csv"), "csv file doesn't exit"
+      hash = {}
+      CSV.foreach("transrate_assemblies.csv", :headers => true,
+                                   :header_converters => :symbol,
+                                   :converters => :all) do |row|
+        row.headers
+        row.fields
+        row.headers.zip(row.fields).each do |header, field|
+          hash[header]=field
+        end
+      end
+      assert_equal 10331, hash[:n_bases], "number of bases"
+      assert_equal 1566, hash[:n50], "n50"
+      assert_equal 10, hash[:n_refs_with_crbb], "number of crb hits"
+    end
+
+    should "fail when one of multiple assemblies is missing" do
+      assembly = File.join(File.dirname(__FILE__), 'data', 'assembly.2.fa')
+      assembly2 = File.join(File.dirname(__FILE__), 'data', 'foo.fa')
+      reference = File.join(File.dirname(__FILE__), 'data', 'Os.protein.2.fa')
+      left = File.join(File.dirname(__FILE__), 'data', '150uncovered.l.fq')
+      right = File.join(File.dirname(__FILE__), 'data', '150uncovered.r.fq')
+      cmd = "bundle exec bin/transrate "
+      cmd << " --assembly #{assembly},#{assembly2}"
+      cmd << " --reference #{reference}"
+      cmd << " --left #{left}"
+      cmd << " --right #{right}"
+      c = Transrate::Cmd.new("#{cmd}")
+      c.run
+      assert_equal false, c.status.success?, "exit status"
+    end
+
+  end
+end

data/test/test_bowtie.rb

@@ -50,5 +50,17 @@ class TestBowtie < Test::Unit::TestCase
         end
       end
     end
+
+    should "raise error when bowtie fails" do
+      not_reads = File.join(File.dirname(__FILE__), 'data', 'not_a_file.fq')
+      Dir.mktmpdir do |tmpdir|
+        Dir.chdir tmpdir do
+          assert_raise Transrate::Bowtie2Error do
+            @mapper.build_index @reference
+            @mapper.map_reads(@reference, @left, not_reads)
+          end
+        end
+      end
+    end
   end
 end

data/test/test_comp_metrics.rb

@@ -1,5 +1,41 @@
 require 'helper'
 
+module CRB_Blast
+  class CRB_Blast
+    def change_hit(query_name, target_name, qstart, qend, tstart, tend, qlen, tlen)
+      hits = @reciprocals[query_name]
+      hits.each do |hit|
+        if hit.target == target_name
+          hit.qstart = qstart
+          hit.qend = qend
+          hit.tstart = tstart
+          hit.tend = tend
+          hit.qlen = qlen
+          hit.tlen = tlen
+        end
+      end
+    end
+
+    def add_hit(query_name, target_name, qstart, qend, tstart, tend, qlen, tlen)
+      @reciprocals[query_name] ||= []
+      list = Array.new(14)
+      list[0] = query_name
+      list[1] = target_name
+      list[6] = qstart
+      list[7] = qend
+      list[8] = tstart
+      list[9] = tend
+      list[12] = qlen
+      list[13] = tlen
+      @reciprocals[query_name] << Hit.new(list)
+    end
+
+    def remove_hit(query_name)
+      @reciprocals.delete(query_name)
+    end
+  end
+end
+
 class TestCompMetrics < Test::Unit::TestCase
 
   context "ComparativeMetrics" do
@@ -11,13 +47,14 @@ class TestCompMetrics < Test::Unit::TestCase
       targetpath = File.join(File.dirname(__FILE__),
                             'data',
                             'Os.protein.2.fa')
-      assembly = Transrate::Assembly.new(querypath)
-      reference = Transrate::Assembly.new(targetpath)
+      @assembly = Transrate::Assembly.new(querypath)
+      @q_ids = @assembly.assembly.keys
+      @reference = Transrate::Assembly.new(targetpath)
+      @t_ids = @reference.assembly.keys
       threads = 8
-      @comp = Transrate::ComparativeMetrics.new(assembly, reference, threads)
+      @comp = Transrate::ComparativeMetrics.new(@assembly, @reference, threads)
     end
 
-
     should "run metrics on assembly" do
       Dir.mktmpdir do |tmpdir|
         Dir.chdir tmpdir do
@@ -27,82 +64,127 @@ class TestCompMetrics < Test::Unit::TestCase
       end
     end
 
-    should "calculate ortholog hit ratio" do
-      crb = CRBHelper.new(false)
-
-      hash = Hash.new
-      (1..11).each do |i|
-        hash["q#{i}"] = []
-      end
-      # Q1   |------------|
-      # Q2                    |--------------|
-      # T1 |------------------------------------|  # coverage = 200/500
-      hash["q1"] << HitHelper.new("q1", "t1", 1, 100, 101, 200, 100, 500)
-      hash["q2"] << HitHelper.new("q2", "t1", 1, 100, 301, 400, 100, 500)
-
-      # Q3                    |------------|
-      # Q4   |--------------|
-      # T2 |------------------------------------| # coverage = 200/500
-      hash["q3"] << HitHelper.new("q3", "t2", 1, 100, 301, 400, 100, 500)
-      hash["q4"] << HitHelper.new("q4", "t2", 1, 100, 101, 200, 100, 500)
-
-      # Q5                |------------|
-      # Q6   |-------------------|
-      # T3 |------------------------------------| # coverage = 300/500
-      hash["q5"] << HitHelper.new("q5", "t3", 1, 200, 201, 400, 200, 500)
-      hash["q6"] << HitHelper.new("q6", "t3", 1, 200, 101, 300, 200, 500)
-
-      # Q7             |------------|
-      # Q8      |------------------------|
-      # T3 |------------------------------------| # coverage = 300/500
-      hash["q7"] << HitHelper.new("q7", "t4", 1, 100, 201, 300, 100, 500)
-      hash["q8"] << HitHelper.new("q8", "t4", 1, 300, 101, 400, 300, 500)
-
-      # Q9     |--------|
-      # Q10                        |--------|
-      # Q11        |--------------------|
-      # T5 |------------------------------------| # coverage = 600/1000
-      hash["q9"] << HitHelper.new("q9",   "t5", 1, 200, 201, 400, 200, 1000)
-      hash["q10"] << HitHelper.new("q10", "t5", 1, 200, 601, 800, 200, 1000)
-      hash["q11"] << HitHelper.new("q11", "t5", 1, 400, 301, 700, 400, 1000)
-
-      crb.hash = hash
-      ohr = @comp.ortholog_hit_ratio crb
-      assert_equal 16.0/30.0, ohr
+    should "calculate reference coverage" do
+      crb = @comp.reciprocal_best_blast
+      # change the results so i know what i have
+      # qstart, qend, tstart, tend, qlen, tlen
+      #
+      # Q       |------------|
+      # T1 |-------------------------|
+      crb.change_hit("scaf_Os03g60760.1", "LOC_Os03g60760.1",
+                     1, 300, 101, 200, 300, 200) # 0.5
+      @reference["LOC_Os03g60760.1"].seq = "A"*200
+      #
+      # Q1   |----------|
+      # Q2                 |------------|
+      # T2  |-------------------------------|
+      crb.change_hit("scaf_Os10g39590.1", "LOC_Os10g39590.1",
+                     1, 150, 51, 100, 150, 200) # 0.25
+      crb.add_hit("scaf_Os10g39590.1", "LOC_Os10g39590.1",
+                     1, 150, 151, 200, 150, 200) # 0.25
+      @reference["LOC_Os10g39590.1"].seq = "A"*200
+      #
+      # adding first block [151..300]  scaf_Os09g38670.1
+      #    450 / 600.0
+      #    LOC_Os09g38670.1  0.75
+
+      #
+      #
+      # Q1           |-----------|
+      # Q2      |----------------------|
+      # T3  |-------------------------------|
+      crb.change_hit("scaf_Os09g38670.1", "LOC_Os09g38670.1",
+                     1, 150, 51, 100, 150, 200) # 0.25
+      crb.add_hit("scaf_Os09g38670.1", "LOC_Os09g38670.1",
+                     1, 450, 26, 175, 450, 200) # 0.75
+      @reference["LOC_Os09g38670.1"].seq = "A"*200
+
+      #
+      # Q1      |----------------------|
+      # Q2           |-----------|
+      # T4  |-------------------------------|
+      crb.change_hit("scaf_Os12g21920.1", "LOC_Os12g21920.1", #
+                     1, 450, 26, 175, 450, 200) # 0.75
+      crb.add_hit("scaf_Os12g21920.1", "LOC_Os12g21920.1",
+                     1, 150, 51, 100, 150, 200) # 0.25
+      @reference["LOC_Os12g21920.1"].seq = "A"*200
+
+
+      #
+      # Q1    |------|
+      # Q2                      |--------|
+      # Q3       |-----------------|
+      # T5  |-------------------------------|
+      crb.change_hit("scaf_Os01g36294.1", "LOC_Os01g36294.1", #
+                     1, 300, 51, 100, 300, 400)
+      crb.add_hit("scaf_Os01g36294.1", "LOC_Os01g36294.1",
+                     1, 300, 200, 250, 300, 400)
+      crb.add_hit("scaf_Os01g36294.1", "LOC_Os01g36294.1",
+                     1, 300, 75, 225, 300, 400)
+      @reference["LOC_Os01g36294.1"].seq = "A"*400
+
+      crb.change_hit("scaf_Os12g22750.1", "LOC_Os12g22750.1",
+                     1, 300, 101, 200, 300, 200) # 0.5 # 300/600
+      @reference["LOC_Os12g22750.1"].seq = "A"*200
+
+      crb.change_hit("scaf_Os02g55190.1", "LOC_Os02g55190.1",
+                     1, 300, 101, 200, 300, 200) # 0.5 # 300/600
+      @reference["LOC_Os02g55190.1"].seq = "A"*200
+
+      crb.change_hit("scaf_Os03g56500.1", "LOC_Os03g56500.1",
+                     1, 300, 101, 200, 300, 400) # 0.25
+      crb.change_hit("scaf_Os03g56500.2", "LOC_Os03g56500.1",
+                     1, 300, 201, 300, 300, 400) # 0.25 # 600 / 1200
+      @reference["LOC_Os03g56500.1"].seq = "A"*400
+
+      crb.change_hit("scaf_Os03g56724.1", "LOC_Os03g56724.1",
+                     1, 300, 101, 200, 300, 200) # 300/600 = 0.5
+      @reference["LOC_Os03g56724.1"].seq = "A"*200
+
+      crb.remove_hit("scaf_Os01g11360.1")
+
+      @reference["LOC_Os03g08270.3"].seq = "A"*200
+      @reference["LOC_Os10g41970.1"].seq = "A"*200
+      @reference["LOC_Os09g26780.1"].seq = "A"*200
+      @reference["LOC_Os12g24659.1"].seq = "A"*200
+      @reference["LOC_Os01g36410.1"].seq = "A"*200
+      @reference["LOC_Os12g22780.1"].seq = "A"*200
+      @reference["LOC_Os02g56470.1"].seq = "A"*200
+      @reference["LOC_Os03g30530.1"].seq = "A"*200
+      @reference["LOC_Os03g49850.1"].seq = "A"*200
+      @reference["LOC_Os01g11360.1"].seq = "A"*200
+      @reference["LOC_Os01g44140.1"].seq = "A"*200
+
+      # total_length of references should be 4400
+
+      cov = @comp.coverage crb
+      assert_equal 3600/4400.0, cov, "reference coverage"
     end
 
     should "calculate potential chimera count" do
-      crb = CRBHelper.new(false)
+      crb = @comp.reciprocal_best_blast
+      # # T1   |---------|
+      # # T2                 |---------|
+      # # Q1 |----------------------------| # chimera = true
+
+      crb.remove_hit("scaf_Os10g39590.1")
+      crb.add_hit("scaf_Os10g39590.1", "LOC_Os03g60760.1",
+                     1, 150, 51, 100, 400, 60) # 0.25
+      crb.add_hit("scaf_Os10g39590.1", "LOC_Os10g39590.1",
+                     200, 350, 51, 100, 400, 60) # 0.25
+
+      # # T3   |---------|
+      # # T3                 |---------|
+      # # Q2 |----------------------------|
+      # # chimera = true because the reference has the region 1-100 duplicated
+      crb.remove_hit("scaf_Os12g21920.1")
+      crb.add_hit("scaf_Os12g21920.1", "LOC_Os12g21920.1",
+                     1, 150, 55, 105, 400, 60) # 0.25
+      crb.add_hit("scaf_Os12g21920.1", "LOC_Os12g21920.1",
+                     200, 350, 51, 100, 400, 60) # 0.25
 
-      hash = Hash.new
-      (1..3).each do |i|
-        hash["q#{i}"] = []
-      end
-
-      # T1   |---------|
-      # T2                 |---------|
-      # Q1 |----------------------------| # chimera = true
-      hash["q1"] << HitHelper.new("q1", "t1", 101, 200, 1, 100, 500, 100)
-      hash["q1"] << HitHelper.new("q1", "t2", 301, 400, 1, 100, 400, 100)
-
-
-      # T3   |---------|
-      # T3                 |---------|
-      # Q2 |----------------------------|
-      # chimera = true because the reference has the region 1-100 duplicated
-      hash["q2"] << HitHelper.new("q2", "t3", 101, 200, 1, 100, 500, 100)
-      hash["q2"] << HitHelper.new("q2", "t3", 301, 400, 1, 100, 400, 100)
-
-      # # T4   |---------|
-      # # T4                 |---------|
-      # # Q3 |----------------------------|
-      # # chimera = false because the reference
-      hash["q3"] << HitHelper.new("q3", "t4", 101, 200, 1, 100, 500, 200)
-      hash["q3"] << HitHelper.new("q3", "t4", 301, 400, 101, 200, 400, 200)
-
-      crb.hash = hash
       @comp.chimeras crb
-      assert_equal 0.667, @comp.p_chimeras.round(3)
+      assert_equal 2/11.0, @comp.p_chimeras
     end
 
     should "calculate overlap amount" do
@@ -116,8 +198,8 @@ class TestCompMetrics < Test::Unit::TestCase
       Dir.mktmpdir do |tmpdir|
         Dir.chdir tmpdir do
           @comp.run
-          assert_equal 11, @comp.comp_stats[:n_contigs_with_recip]
-          assert_equal 11/13.0, @comp.comp_stats[:p_contigs_with_recip]
+          assert_equal 11, @comp.comp_stats[:n_contigs_with_CRBB]
+          assert_equal 11/13.0, @comp.comp_stats[:p_contigs_with_CRBB]
         end
       end
     end
@@ -126,8 +208,8 @@ class TestCompMetrics < Test::Unit::TestCase
       Dir.mktmpdir do |tmpdir|
         Dir.chdir tmpdir do
           @comp.run
-          assert_equal 10, @comp.comp_stats[:n_refs_with_recip]
-          assert_equal 0.5, @comp.comp_stats[:p_refs_with_recip]
+          assert_equal 10, @comp.comp_stats[:n_refs_with_CRBB]
+          assert_equal 0.5, @comp.comp_stats[:p_refs_with_CRBB]
         end
       end
     end
@@ -148,30 +230,5 @@ class TestCompMetrics < Test::Unit::TestCase
       end
     end
 
-    should "number of reference sequences coverage" do
-      # n&p of reference sequences covered to (25, 50, 75, 85, 95%)
-      # of their length by CRB-BLAST hit
-      crb = CRBHelper.new(false)
-
-      hash = Hash.new
-      (1..5).each do |i|
-        hash["q#{i}"] = []
-      end
-      hash["q1"] << HitHelper.new("q1", "t1", 1, 250, 101, 350, 250, 1000)
-      hash["q2"] << HitHelper.new("q2", "t2", 1, 500, 101, 600, 500, 1000)
-      hash["q3"] << HitHelper.new("q3", "t3", 1, 750, 101, 850, 750, 1000)
-      hash["q4"] << HitHelper.new("q4", "t4", 1, 850, 101, 950, 850, 1000)
-      hash["q5"] << HitHelper.new("q5", "t5", 1, 950, 1, 950, 950, 1000)
-
-      crb.hash = hash
-      ohr = @comp.ortholog_hit_ratio crb
-      stats = @comp.comp_stats
-      assert_equal 5, stats[:cov25]
-      assert_equal 4, stats[:cov50]
-      assert_equal 3, stats[:cov75]
-      assert_equal 2, stats[:cov85]
-      assert_equal 1, stats[:cov95]
-    end
-
   end
 end