tree_clusters 0.4.0 → 0.5.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: ea8e2199383e573da42d0a6d265b446de07bea4c
-  data.tar.gz: 775d0ace4a7398438ed4c4a34504ae9d7128d7c1
+  metadata.gz: 6fbd972a3068d716241676c936e3867083ad419a
+  data.tar.gz: 973608883e2ca684541f905cbf6644443e217945
 SHA512:
-  metadata.gz: db7302cad104d6ae05e10a4f02b1880016ba9a960cb44209d283cf9c5d730dc39ebd26a6cf64d40bd0e516d6253692112615bde41dc0e1e935ee0bc582cb36c4
-  data.tar.gz: 8dbb10b889de0f3c7db92d2138c89bdcf5357279daadaf12c97ac2e7d9b74980e3e9219308e4bb3252c93bb92b9230d77e5dc166ebd5633f36e949a54c72ca94
+  metadata.gz: 5e55333d9d293b13d2c9c260962adfd0e9ff8b12b6989a4b4bc6f6a8cc67bced83f843d19e3b751b9e917117c724d6307107403a83aba06f9dfe7c8b30ca6b9b
+  data.tar.gz: 7740ace88a99fd9c424a6d49fb83457b4d79e0295cf08fd0664123d08314bcdaa97391bf4f529613ea3151c44567939301343868e4674f9460445a657d445cec

data/exe/snazzy_clades_key_cols

@@ -8,21 +8,10 @@ require "parse_fasta"
 require "shannon"
 require "fileutils"
 
-def get_low_ent_cols leaves, leaf2attrs, entropy_cutoff
-  low_ent_cols = []
-  alns = leaf2attrs.attrs leaves, :aln
-  aln_cols = alns.transpose
-
-  aln_cols.each_with_index do |aln_col, aln_col_idx|
-    has_gaps = aln_col.any? { |aa| aa == "-" }
-    low_entropy = Shannon::entropy(aln_col.join) <= entropy_cutoff
-
-    if !has_gaps && low_entropy
-      low_ent_cols << (aln_col_idx + 1)
-    end
-  end
+TreeClusters.extend TreeClusters
 
-  Set.new low_ent_cols
+def puts_info outf, clade_id, key_cols
+  outf.puts [clade_id, key_cols.count, key_cols.to_a].join "\t"
 end
 
 opts = Trollop.options do
@@ -62,44 +51,61 @@ opts = Trollop.options do
       default: "snazzy_clades")
 end
 
-FileUtils.mkdir_p opts[:outdir]
+abort_if opts[:tree].nil?,
+         "--tree is a required arg"
+abort_if opts[:mapping].nil?,
+         "--mapping is a required arg"
+abort_if opts[:aln].nil?,
+         "--aln is a required arg"
 
-TreeClusters.extend TreeClusters
-
-tree          = NewickTree.fromFile opts[:tree]
-metadata      = TreeClusters.read_mapping_file opts[:mapping]
-snazzy_clades = TreeClusters.snazzy_clades tree, metadata
+abort_unless_file_exists opts[:tree]
+abort_unless_file_exists opts[:mapping]
+abort_unless_file_exists opts[:aln]
 
-aln_len = nil
-leaf2attrs = TreeClusters::Attrs.new
-ParseFasta::SeqFile.open(opts[:aln]).each_record do |rec|
-  leaf2attrs[rec.id] = { aln: rec.seq.chars }
+TreeClusters.check_ids opts[:tree], opts[:mapping], opts[:aln]
 
-  aln_len ||= rec.seq.length
+abort_unless opts[:entropy_cutoff] >= 0,
+             "--entropy-cutoff must be >= 0"
+abort_unless opts[:clade_size_cutoff] >= 1,
+             "--clade-size-cutoff must be >= 1"
 
-  abort_unless aln_len == rec.seq.length,
-               "Aln len mismatch for #{rec.id}"
-end
+FileUtils.mkdir_p opts[:outdir]
 
+tree          = NewickTree.fromFile opts[:tree]
+metadata      = TreeClusters.read_mapping_file opts[:mapping]
+snazzy_clades = TreeClusters.snazzy_clades tree, metadata
+leaf2attrs    = TreeClusters.read_alignment opts[:aln]
+
+clades_fname  =
+  File.join opts[:outdir],
+            "#{opts[:base]}.snazzy_clades.txt"
+members_fname =
+  File.join opts[:outdir],
+            "#{opts[:base]}.snazzy_clades_clade_members.txt"
+all_key_cols_fname =
+  File.join opts[:outdir],
+            "#{opts[:base]}.snazzy_clades_key_cols.txt"
+key_cols_fname =
+  File.join opts[:outdir],
+            "#{opts[:base]}.snazzy_clades_key_cols.txt"
+key_cols_minus_parent_cols_fname =
+  File.join opts[:outdir],
+            "#{opts[:base]}.snazzy_clades_key_cols_minus_parent_cols.txt"
+key_cols_minus_sibling_cols_fname =
+  File.join opts[:outdir],
+            "#{opts[:base]}.snazzy_clades_key_cols_minus_sibling_cols.txt"
+
+info_f =
+  File.open(clades_fname, "w")
+clade_members_f =
+  File.open(members_fname, "w")
+key_cols_f =
+  File.open(key_cols_fname, "w")
+key_cols_minus_parent_cols_f =
+  File.open(key_cols_minus_parent_cols_fname, "w")
+key_cols_minus_sibling_cols_f =
+  File.open(key_cols_minus_sibling_cols_fname, "w")
 
-clades_fname  = File.join opts[:outdir],
-                          "#{opts[:base]}.snazzy_clades.txt"
-members_fname = File.join opts[:outdir],
-                          "#{opts[:base]}.snazzy_clades_clade_members.txt"
-all_key_cols_fname = File.join opts[:outdir],
-                               "#{opts[:base]}.snazzy_clades_key_cols.txt"
-key_cols_fname = File.join opts[:outdir],
-                               "#{opts[:base]}.snazzy_clades_key_cols.txt"
-key_cols_minus_parent_cols_fname = File.join opts[:outdir],
-                               "#{opts[:base]}.snazzy_clades_key_cols_minus_parent_cols.txt"
-key_cols_minus_sibling_cols_fname = File.join opts[:outdir],
-                               "#{opts[:base]}.snazzy_clades_key_cols_minus_sibling_cols.txt"
-
-info_f                        = File.open(clades_fname, "w")
-clade_members_f               = File.open(members_fname, "w")
-key_cols_f                    = File.open(key_cols_fname, "w")
-key_cols_minus_parent_cols_f  = File.open(key_cols_minus_parent_cols_fname, "w")
-key_cols_minus_sibling_cols_f = File.open(key_cols_minus_sibling_cols_fname, "w")
 
 begin
   # info is { metadata_category => metadata_tag , ... }
@@ -115,27 +121,34 @@ begin
                           clade.all_leaves].join "\t"
 
     key_cols_all_leaves =
-      get_low_ent_cols clade.all_leaves, leaf2attrs, opts[:entropy_cutoff]
+      TreeClusters.low_ent_cols clade.all_leaves,
+                                leaf2attrs,
+                                opts[:entropy_cutoff]
     key_cols_all_sibling_leaves =
-      get_low_ent_cols clade.all_sibling_leaves, leaf2attrs, opts[:entropy_cutoff]
+      TreeClusters.low_ent_cols clade.all_sibling_leaves,
+                                leaf2attrs,
+                                opts[:entropy_cutoff]
     key_cols_parent_leaves =
-      get_low_ent_cols clade.parent_leaves, leaf2attrs, opts[:entropy_cutoff]
+      TreeClusters.low_ent_cols clade.parent_leaves,
+                                leaf2attrs,
+                                opts[:entropy_cutoff]
 
     key_cols_all_minus_sibling =
       key_cols_all_leaves - key_cols_all_sibling_leaves
     key_cols_all_minus_parent =
       key_cols_all_leaves - key_cols_parent_leaves
 
-    key_cols_f.puts [clade_id,
-                     key_cols_all_leaves.count,
-                     key_cols_all_leaves.to_a].join "\t"
-    key_cols_minus_parent_cols_f.puts [clade_id,
-                                       key_cols_all_minus_parent.count,
-                                       key_cols_all_minus_parent.to_a].join "\t"
-    key_cols_minus_sibling_cols_f.puts [clade_id,
-                                        key_cols_all_minus_sibling.count,
-                                        key_cols_all_minus_sibling.to_a].join "\t"
+    puts_info key_cols_f,
+              clade_id,
+              key_cols_all_leaves
+
+    puts_info key_cols_minus_parent_cols_f,
+              clade_id,
+              key_cols_all_minus_parent
 
+    puts_info key_cols_minus_sibling_cols_f,
+              clade_id,
+              key_cols_all_minus_sibling
   end
 ensure
   info_f.close

data/lib/tree_clusters.rb

@@ -1,6 +1,8 @@
 require "abort_if"
 require "Newick"
 require "set"
+require "parse_fasta"
+require "shannon"
 require "tree_clusters/version"
 
 include AbortIf
@@ -25,6 +27,96 @@ end
 # Top level namespace of the Gem.
 module TreeClusters
 
+  # Given an ary of strings, find the most common string in the ary.
+  #
+  # @param bases [Array<String>] an array of strings
+  #
+  # @return most_common_str [String] the most common string in the ary.
+  #
+  # @example Upper case and lower case count as the same.
+  #   TreeClusters::consensus %w[a A C T] #=> "A"
+  # @example Ties take the one closest to the end
+  #   TreeClusters::consensus %w[a c T t C t g] #=> "T"
+  #
+  # @note Each string is upcase'd before frequencies are calculated.
+  def consensus bases
+    bases.
+      map(&:upcase).
+      group_by(&:itself).
+      sort_by { |_, bases| bases.count }.
+      reverse.
+      first.
+      first
+  end
+
+  def read_alignment aln_fname
+    leaf2attrs = TreeClusters::Attrs.new
+    aln_len = nil
+    ParseFasta::SeqFile.open(aln_fname).each_record do |rec|
+      leaf2attrs[rec.id] = { aln: rec.seq.chars }
+
+      aln_len ||= rec.seq.length
+
+      abort_unless aln_len == rec.seq.length,
+                   "Aln len mismatch for #{rec.id}"
+    end
+
+    leaf2attrs
+  end
+
+  def low_ent_cols leaves, leaf2attrs, entropy_cutoff
+    low_ent_cols = []
+    alns = leaf2attrs.attrs leaves, :aln
+    aln_cols = alns.transpose
+
+    aln_cols.each_with_index do |aln_col, aln_col_idx|
+      has_gaps = aln_col.any? { |aa| aa == "-" }
+      low_entropy =
+        Shannon::entropy(aln_col.join.upcase) <= entropy_cutoff
+
+      if !has_gaps && low_entropy
+        low_ent_cols << (aln_col_idx + 1)
+      end
+    end
+
+    Set.new low_ent_cols
+  end
+
+  def check_ids tree, mapping, aln
+    tree_ids = Set.new(NewickTree.fromFile(tree).taxa)
+
+    mapping_ids = Set.new
+    File.open(mapping, "rt").each_line.with_index do |line, idx|
+      unless idx.zero?
+        id, *rest = line.chomp.split
+
+        mapping_ids << id
+      end
+    end
+
+    aln_ids = Set.new
+    ParseFasta::SeqFile.open(aln).each_record do |rec|
+      aln_ids << rec.id
+    end
+
+    if !(tree_ids == mapping_ids && mapping_ids == aln_ids)
+      AbortIf::logger.error { "Seq IDs did not match in all input files" }
+
+      tree_ids = tree_ids.to_a.sort
+      mapping_ids = mapping_ids.to_a.sort
+      aln_ids = aln_ids.to_a.sort
+
+      AbortIf::logger.debug { ["tree_ids", tree_ids].join "\t" }
+      AbortIf::logger.debug { ["mapping_ids", mapping_ids].join "\t" }
+      AbortIf::logger.debug { ["aln_ids", aln_ids].join "\t" }
+
+      raise AbortIf::Exit
+    else
+      true
+    end
+  end
+
+
   # Given a NewickTree, return an array of all Clades in that tree.
   #
   # @param tree [NewickTree] a NewickTree object

data/lib/tree_clusters/version.rb

@@ -1,3 +1,3 @@
 module TreeClusters
-  VERSION = "0.4.0"
+  VERSION = "0.5.0"
 end

data/test_files/bad.aln

@@ -0,0 +1,6 @@
+>s1
+AAA
+>s2
+AAAA
+>s3
+AA

data/test_files/small.aln

@@ -1,14 +1,14 @@
->a-1
+>a-1 apple
 AAAA
->a-2
+>a-2 pie
 AAAT
->b-1
+>b-1 is
 CCCC
->b-2
-CCCT
->bb-1
+>b-2 really
+cccT
+>bb-1 good
 CCTG
->bbb-1
-CCGG
->bbb-2
+>bbb-1 and
+CCGg
+>bbb-2 tasty
 CGGG

data/test_files/small_aln_bad_ids

@@ -0,0 +1,4 @@
+>apple
+aRSOIN
+>pie
+ARSOIT

data/test_files/small_mapping_bad_ids

@@ -0,0 +1,2 @@
+snazzy	pie
+apple   thing

data/test_files/small_tree_bad_ids

@@ -0,0 +1 @@
+(a, b);

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: tree_clusters
 version: !ruby/object:Gem::Version
-  version: 0.4.0
+  version: 0.5.0
 platform: ruby
 authors:
 - Ryan Moore
@@ -191,11 +191,15 @@ files:
 - exe/snazzy_clades_key_cols
 - lib/tree_clusters.rb
 - lib/tree_clusters/version.rb
+- test_files/bad.aln
 - test_files/non_bifurcating.aln
 - test_files/non_bifurcating.tre
 - test_files/small.aln
 - test_files/small.mapping
 - test_files/small.tre
+- test_files/small_aln_bad_ids
+- test_files/small_mapping_bad_ids
+- test_files/small_tree_bad_ids
 - test_files/test.tre
 - tree_clusters.gemspec
 homepage: https://github.com/mooreryan/tree_clusters