tree_clusters 0.8.2 → 0.8.3

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 7f7ba7426f7048a684ba53a52611efe18cda9603
-  data.tar.gz: c511525c40d1b766c08c2a58e6076633d179439c
+  metadata.gz: be727e6384cf510290ddab3bf5aabd5075505e77
+  data.tar.gz: 69e94311ac35dcb26579edada3d6555b56048cd1
 SHA512:
-  metadata.gz: 4f5ad29eb52cbcdf997a7d1f2431bf9f26adf76c534bd6bbc74fc79c05b331f224d7e8b056173e4b49454a324ee5eb14c2f8dc401379552a8b9fed36627e0eff
-  data.tar.gz: 8d7885844452c979abdfbc71ee85c4ecedefa8818047062ede26eb894bc1913ae612093e07cf68c760aa12654e448b97ce323d6982f5589e2f13821debbf3bd7
+  metadata.gz: 5aacaa9b69741eff4f23b62b7be230befc444847cbf869ed0a5394e1fd768b34fc150483a71f079aef6fd683a9c58fe170e8ffdfc6f229b677862ce7b851c875
+  data.tar.gz: c3e639c1bced1b5c480e99f0d9e19ecb65a27dc3cf09e81d429714951d47b7b66905b82b9ed20d14ac05d36be0067b9b3de51796447acfd91a814ce73ed07fdc

data/.gitignore

@@ -31,5 +31,6 @@ todo.txt
 
 .idea
 ARST
+RNR
 
 exe/key_cols2

data/exe/key_cols

@@ -1,5 +1,7 @@
 #!/usr/bin/env ruby
 
+# TODO the block variables called "clade" aren't actually instances of the TreeClusters::Clade class.  It's kind of confusing.
+
 Signal.trap("PIPE", "EXIT")
 
 require "tree_clusters"
@@ -54,6 +56,8 @@ The key_cols program
   alignment file (fasta format), I will tell you key columns for all
   clades/clusters that have them.
 
+  Version: v#{TreeClusters::VERSION}
+
 Overview
 --------
 
@@ -147,9 +151,10 @@ abort_unless opts[:clade_size_cutoff] >= 1,
 
 FileUtils.mkdir_p opts[:outdir]
 
-AbortIf.logger.info { "Parsing input files" }
+AbortIf.logger.info { "Reading tree" }
+tree = NewickTree.fromFile opts[:tree]
 
-tree       = NewickTree.fromFile opts[:tree]
+AbortIf.logger.info { "Reading alignment" }
 leaf2attrs = TreeClusters.read_alignment opts[:aln]
 
 members_fname        =
@@ -169,22 +174,37 @@ key_cols_f       =
 annotated_tree_f =
     File.open(annotated_tree_fname, "w")
 
-key_col_sets       = {}
-clade_sizes        = {}
-clade_count        = TreeClusters.all_clades(tree).count
+key_col_sets = {}
+clade_sizes  = {}
+# AbortIf.logger.info { "Counting clades" }
+# clade_count        = TreeClusters.all_clades(tree).count
 change_these_names = Set.new
 
 def all_clades_helper tree
-  TreeClusters.all_clades(tree).sort_by { |cl| cl.all_leaves.count }.reverse
+  # TreeClusters.all_clades(tree).sort_by { |cl| cl.all_leaves.count }.reverse
+  tree.clade_nodes.
+      reverse.
+      sort_by { |node| node.all_leaves.count }.
+      reverse
 end
 
-AbortIf.logger.info { "Processing clades" }
+AbortIf.logger.info { "Getting all clades" }
+clades      = all_clades_helper(tree)
+clade_count = clades.count
+
+AbortIf.logger.info { "Processing clades (The first few clades go reaaally slowly, but then I speed up!)" }
 begin
-  all_clades_helper(tree).each_with_index do |clade, idx|
-    if ((idx + 1) % 100).zero?
+  clades.each_with_index do |clade, idx|
+    # It starts off really slowly, then speeds up a lot.
+    if (idx+1) < 100 ||
+        ((idx+1) < 1000 && ((idx + 1) % 10).zero?) ||
+        ((idx+1) < 10000 && ((idx + 1) % 100).zero?) ||
+        ((idx+1) < 100000 && ((idx + 1) % 1000).zero?) ||
+        ((idx + 1) % 10000).zero?
+
       perc = ((idx + 1) / clade_count.to_f * 100).round 2
 
-      STDERR.printf("Processing clades: #{perc}%\r")
+      STDERR.printf("Processing clades: #{idx + 1} of #{clade_count} (#{perc}%%)\r")
     end
 
     clade_id              = "clade_#{idx + 1}___#{clade.name.tr("'", "_")}"
@@ -224,12 +244,12 @@ begin
 
   AbortIf.logger.info { "Annotating tree" }
 
-  all_clades_helper(tree).each_with_index do |clade, idx|
+  clades.each_with_index do |clade, idx|
     clade_id = "clade_#{idx + 1}___#{clade.name.tr("'", "_")}"
 
     if change_these_names.include? clade_id
       # This will change the node in the original NewickTree
-      clade.node.name = "'#{clade_id}'"
+      clade.name = "'#{clade_id}'"
     end
   end
 

data/lib/tree_clusters.rb

@@ -27,12 +27,20 @@ class NewickTree
   end
 
   def unquoted_taxa
+    # @note self.taxa calls taxa method on the root of the tree.
     self.taxa.map { |str| str.tr %q{"'}, "" }
   end
 end
 
+class NewickNode
+  def all_leaves
+    self.leaves.map { |n| n.name.tr  %q{"'}, "" }
+  end
+end
+
 # Top level namespace of the Gem.
 module TreeClusters
+  PROJ_ROOT = File.join __dir__, ".."
 
   # Given an ary of strings, find the most common string in the ary.
   #
@@ -48,18 +56,24 @@ module TreeClusters
   # @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
+        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
+    aln_len    = nil
+    seq_num = 0
     ParseFasta::SeqFile.open(aln_fname).each_record do |rec|
+      seq_num += 1
+      if ((seq_num + 1) % 1000).zero?
+        STDERR.printf("Reading alignment: #{seq_num + 1}\r")
+      end
+
       leaf2attrs[rec.id] = { aln: rec.seq.chars }
 
       aln_len ||= rec.seq.length
@@ -73,13 +87,13 @@ module TreeClusters
 
   def low_ent_cols leaves, leaf2attrs, entropy_cutoff
     low_ent_cols = []
-    alns = leaf2attrs.attrs leaves, :aln
-    aln_cols = alns.transpose
+    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 == "-" }
+      has_gaps    = aln_col.any? { |aa| aa == "-" }
       low_entropy =
-        Shannon::entropy(aln_col.join.upcase) <= entropy_cutoff
+          Shannon::entropy(aln_col.join.upcase) <= entropy_cutoff
 
       if !has_gaps && low_entropy
         low_ent_cols << (aln_col_idx + 1)
@@ -92,11 +106,11 @@ module TreeClusters
   # Like low_ent_cols method but also returns the bases at the positions.
   def low_ent_cols_with_bases leaves, leaf2attrs, entropy_cutoff
     low_ent_cols = []
-    alns = leaf2attrs.attrs leaves, :aln
-    aln_cols = alns.transpose
+    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 == "-" }
+      has_gaps    = aln_col.any? { |aa| aa == "-" }
       low_entropy =
           Shannon::entropy(aln_col.join.upcase) <= entropy_cutoff
 
@@ -130,9 +144,9 @@ module TreeClusters
     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
+      tree_ids    = tree_ids.to_a.sort
       mapping_ids = mapping_ids.to_a.sort
-      aln_ids = aln_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" }
@@ -152,7 +166,7 @@ module TreeClusters
   # @yieldparam clade [Clade] a clade of the tree
   #
   # @return [Enumerator<Clade>] enumerator of Clade objects
-  def all_clades tree, metadata=nil
+  def all_clades tree, metadata = nil
     return enum_for(:all_clades, tree, metadata) unless block_given?
 
     tree.clade_nodes.reverse.each do |node|
@@ -164,12 +178,12 @@ module TreeClusters
     snazzy_clades = {}
 
     clades = self.
-             all_clades(tree, metadata).
-             sort_by { |clade| clade.all_leaves.count }.
-             reverse
+        all_clades(tree, metadata).
+        sort_by { |clade| clade.all_leaves.count }.
+        reverse
 
     metadata.each do |md_cat, leaf2mdtag|
-      already_checked = Set.new
+      already_checked   = Set.new
       single_tag_clades = {}
 
       clades.each do |clade|
@@ -177,8 +191,8 @@ module TreeClusters
                "A clade cannot also be a leaf"
 
         unless clade.all_leaves.all? do |leaf|
-                 already_checked.include? leaf
-               end
+          already_checked.include? leaf
+        end
           md_tags = clade.all_leaves.map do |leaf|
             assert leaf2mdtag.has_key?(leaf),
                    "leaf #{leaf} is missing from leaf2mdtag ht"
@@ -224,9 +238,9 @@ module TreeClusters
     snazzy_info = {}
 
     clades = self.
-             all_clades(tree, metadata).
-             sort_by { |clade| clade.all_leaves.count }.
-             reverse
+        all_clades(tree, metadata).
+        sort_by { |clade| clade.all_leaves.count }.
+        reverse
 
     # Non snazzy clades have a value of nil, so set all to nil and the
     # snazzy ones will be overwritten.
@@ -235,7 +249,7 @@ module TreeClusters
     end
 
     metadata.each do |md_cat, leaf2mdtag|
-      already_checked = Set.new
+      already_checked   = Set.new
       single_tag_clades = {}
 
       clades.each do |clade|
@@ -243,8 +257,8 @@ module TreeClusters
                "A clade cannot also be a leaf"
 
         unless clade.all_leaves.all? do |leaf|
-                 already_checked.include? leaf
-               end
+          already_checked.include? leaf
+        end
           md_tags = clade.all_leaves.map do |leaf|
             assert leaf2mdtag.has_key?(leaf),
                    "leaf #{leaf} is missing from leaf2mdtag ht"
@@ -289,7 +303,7 @@ module TreeClusters
 
   def read_mapping_file fname
     md_cat_names = nil
-    metadata = TreeClusters::Attrs.new
+    metadata     = TreeClusters::Attrs.new
 
     File.open(fname, "rt").each_line.with_index do |line, idx|
       leaf_name, *metadata_vals = line.chomp.split "\t"
@@ -342,6 +356,6 @@ module TreeClusters
       end
     end
 
-      [attr_names, attrs]
+    [attr_names, attrs]
   end
 end

data/lib/tree_clusters/version.rb

@@ -1,3 +1,3 @@
 module TreeClusters
-  VERSION = "0.8.2"
+  VERSION = "0.8.3"
 end

data/test_files/key_cols/expected_output/small2/ARST.tree_clusters.annotated_tree.txt

@@ -0,0 +1 @@
+(('a-1','a-2')'clade_3___cluster_A',(("b-1",b-2)'clade_4___cluster_B1',(bb-1,(bbb-1,bbb-2)'clade_5___cluster_B3')'clade_2___cluster_B2')'clade_1___cluster_B')cluster_C;

data/test_files/key_cols/expected_output/small2/ARST.tree_clusters.clade_members.txt

@@ -0,0 +1,5 @@
+clade_1___cluster_B	5	b-1	b-2	bb-1	bbb-1	bbb-2
+clade_2___cluster_B2	3	bb-1	bbb-1	bbb-2
+clade_3___cluster_A	2	a-1	a-2
+clade_4___cluster_B1	2	b-1	b-2
+clade_5___cluster_B3	2	bbb-1	bbb-2

data/test_files/key_cols/expected_output/small2/ARST.tree_clusters.key_cols.txt

@@ -0,0 +1,5 @@
+clade_1___cluster_B	1	1-C
+clade_2___cluster_B2	2	1-C	4-G
+clade_3___cluster_A	4	1-A	2-A	3-A	5-G
+clade_4___cluster_B1	4	1-C	2-C	3-C	5-A
+clade_5___cluster_B3	4	1-C	3-G	4-G	5-G

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: tree_clusters
 version: !ruby/object:Gem::Version
-  version: 0.8.2
+  version: 0.8.3
 platform: ruby
 authors:
 - Ryan Moore
@@ -201,6 +201,9 @@ files:
 - lib/tree_clusters/clade.rb
 - lib/tree_clusters/version.rb
 - test_files/bad.aln
+- test_files/key_cols/expected_output/small2/ARST.tree_clusters.annotated_tree.txt
+- test_files/key_cols/expected_output/small2/ARST.tree_clusters.clade_members.txt
+- test_files/key_cols/expected_output/small2/ARST.tree_clusters.key_cols.txt
 - test_files/non_bifurcating.aln
 - test_files/non_bifurcating.tre
 - test_files/small.aln