sequence_logo 1.0.2

data/lib/sequence_logo/ytilib/bismark.rb

@@ -0,0 +1,71 @@
+#!/usr/bin/ruby
+
+module Ytilib
+require "rexml/document"
+include REXML
+
+class Bismark < Document
+  
+  def initialize(source = nil, add_dtd = false)
+    dtd = add_dtd ? "<!DOCTYPE smallbismark SYSTEM 'smallbismark.dtd'>#{$/}" : ""
+    source == nil ? super("<?xml version='1.0' encoding='UTF-8'?>#{$/}#{dtd}") : super(source)
+    super(IO.read(source)) if source != nil && root == nil
+    if source == nil
+      self.add_element("smallbismark")
+      # xmlns breaks XPath for a REXML library under Linux, strange, indeed
+      # self.add_element("smallbismark", {"xmlns" => "http://bioinform.imb.ac.ru/smallBiSMark/smallbismark.dtd"})
+      self.root.add_element("comment", {"name" => "WARNING"}).add_text("This is a draft version of small-BiSMark. Specification is the subject to change!")
+    end
+  end
+  
+  def getXML
+    beautify
+    s = ""; write(s, 1, true)
+    s.rstrip!
+    return s
+  end
+  alias get_xml getXML
+  
+  def get_pm(xpath)
+    pwmnode = self.elements[xpath]
+    pm = PM.new_pm(pwmnode.attribute("length").value.to_i)
+    toi = pwmnode.name == "PCM"
+    pwmnode.elements.each("pm-column") { |c|
+      position = c.attribute("position").value.to_i - 1
+      weights = [c.elements["a"].get_text.value.strip.to_f, 
+                c.elements["c"].get_text.value.strip.to_f, 
+                c.elements["g"].get_text.value.strip.to_f, 
+                c.elements["t"].get_text.value.strip.to_f]
+      weights.collect { |w| w.to_i } if toi
+      pm['A'][position], pm['C'][position], pm['G'][position], pm['T'][position] = weights[0], weights[1], weights[2], weights[3]
+    }
+    return pm
+  end
+  
+private
+  CONTAIN_NO_TEXT = { 
+                      "segment" => :vasya_shmyak,
+                      "group" => :vasya_shmyak,
+                      "smallbismark" => :vasya_shmyak,
+                      "motif" => :vasya_shmyak,
+                      "PWM" => :vasya_shmyak,
+                      "PCM" => :vasya_shmyak,
+                      "PPM" => :vasya_shmyak,
+                      "source" => :vasya_shmyak,
+                      "factor"  => :vasya_shmyak,
+                      "pm-column" => :vasya_shmyak,
+                      "word-list" => :vasya_shmyak}
+  
+  def beautify(node = self)
+    if node == self 
+      self.delete_if { |e| e.is_a?(Text) }
+      self.each { |e| beautify(e) }
+    else
+      node.delete_if { |e| e.is_a?(Text) } if node.respond_to?(:delete_if) && Bismark::CONTAIN_NO_TEXT.has_key?(node.name)
+      node.each { |e| beautify(e) } if node.respond_to?(:each)
+    end 
+  end
+
+end
+
+end

data/lib/sequence_logo/ytilib/hack1.rb

@@ -0,0 +1,75 @@
+require 'rexml/formatters/pretty'
+
+module REXML
+  module Formatters
+    # The Transitive formatter writes an XML document that parses to an
+    # identical document as the source document.  This means that no extra
+    # whitespace nodes are inserted, and whitespace within text nodes is
+    # preserved.  Within these constraints, the document is pretty-printed,
+    # with whitespace inserted into the metadata to introduce formatting.
+    #
+    # Note that this is only useful if the original XML is not already
+    # formatted.  Since this formatter does not alter whitespace nodes, the
+    # results of formatting already formatted XML will be odd.
+    class Transitive < Default
+      def initialize( indentation=2 )
+        @indentation = indentation
+        @level = 0
+      end
+      
+      protected
+      def write_element( node, output )
+        output << "\n" << ' '*@level
+        output << "<#{node.expanded_name}"
+
+        node.attributes.each_attribute do |attr|
+          output << " "
+          attr.write( output )
+        end unless node.attributes.empty?
+
+        if node.children.empty?
+          output << "/>"
+        else
+          output << ">"
+          # If compact and all children are text, and if the formatted output
+          # is less than the specified width, then try to print everything on
+          # one line
+          skip = false
+          @level += @indentation
+          
+          only_text = true
+          
+          node.children.each { |child|
+          	only_text = child.is_a?(REXML::Text) && only_text
+            write( child, output )
+          }
+          @level -= @indentation
+          output << "#{only_text ? "" : "\n" + ' '*@level}" << "</#{node.expanded_name}>"
+        end
+        
+      end
+
+      def write_text( node, output )
+        output << node.to_s()
+      end
+    end
+  end
+  
+  class Document
+    def write( output=$stdout, indent=-1, trans=false, ie_hack=false )
+      if xml_decl.encoding != "UTF-8" && !output.kind_of?(Output)
+        output = Output.new( output, xml_decl.encoding )
+      end
+      formatter = if indent > -1
+          if trans
+            REXML::Formatters::Transitive.new( indent )
+          else
+            REXML::Formatters::Pretty.new( indent, ie_hack )
+          end
+        else
+          REXML::Formatters::Default.new( ie_hack )
+        end
+      formatter.write( self, output )
+    end
+  end
+end

data/lib/sequence_logo/ytilib/infocod.rb

@@ -0,0 +1,108 @@
+#!/usr/bin/ruby
+
+class Float
+  # Using Stieltjes formula from http://www.luschny.de/math/factorial/approx/SimpleCases.html
+  def log_fact
+    return 0.0 if self <= 1
+    a0 = 1.0/12
+    a1 = 1.0/30 
+    a2 = 53.0/210
+    a3 = 195.0/371
+    a4 = 22999.0/22737
+    a5 = 29944523.0/19733142
+    a6 = 109535241009.0/48264275462
+    z_big = self+1;
+    (1.0/2)*Math.log(2*Math::PI)+(z_big-1.0/2)*Math.log(z_big)-z_big + a0/(z_big+a1/(z_big+a2/(z_big+a3/(z_big+a4/(z_big+a5/(z_big+a6/z_big))))))
+  end
+end
+
+class Integer
+  def log_fact
+    self.to_f.log_fact
+  end
+end
+
+# Naive version
+=begin
+class Integer
+  @@fact_hash = {}
+  def log_fact
+    return 0.0 if self == 0
+    return nil if self < 0
+    if self <= 170
+      @@fact_hash[self] = Math.log( lambda { |k| return k if self.times { |i| k *= i.next } }.call(1) )
+    else
+      return self.to_f.log_fact
+    end unless @@fact_hash.has_key?(self)
+    return @@fact_hash[self] 
+  end
+end
+=end
+
+module Ytilib
+  class PM
+    def infocod(position = nil)
+      return infocod_private(position) if position
+      (0...@size).collect { |i| infocod_private(i) }
+    end
+    alias icd infocod
+    
+    def icd2of4(floor = false)
+      i2o4 = @words_count / 2.0
+      i2o4 = i2o4.floor if floor
+      ([i2o4, i2o4, 0, 0].inject(0.0) { |sum, k_i| sum += k_i.log_fact  } - @words_count.log_fact ) / @words_count
+      # 0 is equal to @words_count % 2, because 0! = 1!
+    end
+    
+    def icd3of4(floor = false)
+      i3o4 = @words_count / 3.0
+      i3o4 = i3o4.floor if floor
+      addon = floor ? @words_count % 3 : 0
+      ([i3o4, i3o4, i3o4, addon].inject(0.0) { |sum, k_i| sum += k_i.log_fact  } - @words_count.log_fact ) / @words_count
+    end
+    
+    def icdThc
+      icd3of4
+    end
+    
+    def icdTlc
+      io = @words_count / 6.0
+      ([2*io, 2*io, io, io].inject(0.0) { |sum, k_i| sum += k_i.log_fact  } - @words_count.log_fact ) / @words_count
+    end
+    
+    def icd4of4(floor = false)
+      i4o4 = @words_count / 4.0
+      i4o4 = i4o4.floor if floor
+      ([i4o4, i4o4, i4o4, i4o4].inject(0.0) { |sum, k_i| sum += k_i.log_fact  } - @words_count.log_fact ) / @words_count
+    end
+    
+  protected
+    def infocod_private(position)
+      k_i = ['A','C','G','T'].collect { |letter| @matrix[letter][position] }  
+      ( k_i.inject(0.0) { |sum, k_i| sum += k_i.log_fact  } - @words_count.log_fact ) / @words_count
+    end
+  end
+  
+  class PPM
+    def to_pcm(words_count = nil)
+      @words_count = words_count if words_count
+      checkerr("words count is not specified") { !@words_count }
+      counts = PM.new_matrix(@size)
+      (0...size).each { |i|
+        ['A', 'C', 'G', 'T'].each { |l|
+          counts[l][i] = @matrix[l][i] * @words_count
+        }
+      }
+      return PM.new(size, counts)
+    end
+    alias to_pcm get_pcm
+    
+    def infocod(position = nil)
+      return to_pcm.infocod(position)
+    end
+    
+    def icd(position = nil)
+      return to_pcm.infocod(position)
+    end
+  end
+end

data/lib/sequence_logo/ytilib/iupac.rb

@@ -0,0 +1,92 @@
+class IUPAC < String
+  CODE = {"A" => "A", "C" => "C", "G" => "G", "T" => "T", 
+          "AG" => "R", "CT" => "Y", "GT" => "K", "AC" => "M", 
+          "CG" => "S", "AT" => "W", "CGT" => "B", "AGT" => "D", "ACT" => "H", "ACG" => "V", "ACGT" => "N"}
+  REVCODE = CODE.invert
+  
+  def dup
+    IUPAC.new(self)
+  end
+  
+  def initialize(words)
+    if words.is_a?(Array)
+      iupac = (0...words[0].size).collect { |i|
+        (0...words.size).collect { |j| words[j][i,1] }.uniq.sort.inject("") { |cola, letter| cola += letter }
+      }.inject("") { |iup, cola|
+        checkerr("bad letter set #{cola}") { !CODE.has_key?(cola) }
+        iup += CODE[cola]
+      }
+      super(iupac)
+    elsif words.is_a?(IUPAC)
+      super(words)
+    elsif words.is_a?(String)
+      checkerr("word #{words} has strange characters") { words.tr('ACGTURYKMSWBDHVN', '').size > 0 }
+      super(words)
+    end
+  end
+  
+  def ==(iupac)
+    return false if self.size != iupac.size
+    (0...self.size).inject(true) { |result, i| result &= IUPACOM[self[i,1]][iupac[i,1]] }
+  end
+  
+  def merge(iupac)
+    return nil if self.size != iupac.size
+    res = (0...self.size).inject("") { |res, i|
+      merges = REVCODE[self[i,1]].split(//).concat(REVCODE[iupac[i,1]].split(//)).uniq.sort.inject("") { |s, c| s += c}
+      res << CODE[merges]
+    }
+    return IUPAC.new(res)
+  end
+  
+  def include?(iupac)
+    return false if self.size < iupac.size || !iupac.is_a?(IUPAC)
+    (0..self.size-iupac.size).each { |i|
+      return i if IUPAC.new(self[i,iupac.size]) == iupac
+    }
+    return false
+  end
+  
+  def compl
+    return self.tr("ACGTRYKMSWBDHVN", "TGCAYRMKSWVHDBN")
+  end
+  
+  def compl!
+    self.tr!("ACGTRYKMSWBDHVN", "TGCAYRMKSWVHDBN")
+    return self
+  end
+  
+  alias reverse_string reverse
+  def reverse
+    return IUPAC.new(reverse_string)
+  end
+  
+  alias comp! compl!
+  alias complement! compl!
+  alias comp compl
+  alias complement compl
+  
+private
+  IUPACOM = { "A" => {"A" => :llib, "R" => :llib, "M" => :llib, "W" => :llib, "D" => :llib, "H" => :llib, "V" => :llib, "N" => :llib},
+                "C" => {"C" => :llib, "Y" => :llib, "M" => :llib, "S" => :llib, "B" => :llib, "H" => :llib, "V" => :llib, "N" => :llib},
+                "G" => {"G" => :llib, "R" => :llib, "K" => :llib, "S" => :llib, "B" => :llib, "D" => :llib, "V" => :llib, "N" => :llib},
+                "T" => {"T" => :llib, "Y" => :llib, "K" => :llib, "W" => :llib, "B" => :llib, "D" => :llib, "H" => :llib, "N" => :llib}
+  }
+  IUPACOM["R"] = IUPACOM["G"].merge(IUPACOM["A"])
+  IUPACOM["Y"] = IUPACOM["T"].merge(IUPACOM["C"])
+  IUPACOM["K"] = IUPACOM["G"].merge(IUPACOM["T"])
+  IUPACOM["M"] = IUPACOM["A"].merge(IUPACOM["C"])
+  IUPACOM["S"] = IUPACOM["G"].merge(IUPACOM["C"])
+  IUPACOM["W"] = IUPACOM["A"].merge(IUPACOM["T"])
+  IUPACOM["B"] = IUPACOM["G"].merge(IUPACOM["T"].merge(IUPACOM["C"]))
+  IUPACOM["D"] = IUPACOM["G"].merge(IUPACOM["A"].merge(IUPACOM["T"]))
+  IUPACOM["H"] = IUPACOM["A"].merge(IUPACOM["C"].merge(IUPACOM["T"]))
+  IUPACOM["V"] = IUPACOM["G"].merge(IUPACOM["C"].merge(IUPACOM["A"]))
+  IUPACOM["N"] = IUPACOM["A"].merge(IUPACOM["C"].merge(IUPACOM["G"].merge(IUPACOM["T"])))
+ 
+#  IUPACMERGE = CODE.merge({
+#    "AA" => "A", "CC" => "C", "GG" => "G", "TT" => "T",
+#    
+#  })
+
+end

data/lib/sequence_logo/ytilib/pm.rb

@@ -0,0 +1,562 @@
+module Ytilib
+  class PM
+
+    attr_reader :matrix, :size
+    attr_accessor :words_count
+    
+    alias length size
+    
+    def score_mean(bckgr = Randoom::DEF_PROBS)
+      (0...@size).inject(0.0) { |mean, i| mean += ['A','C','G','T'].inject(0.0) { |sum,l| sum += @matrix[l][i] * bckgr[l] } }
+    end
+    
+    def score_variance(bckgr = Randoom::DEF_PROBS)
+      (0...@size).inject(0.0) { |m2, i| 
+        deltai = ['A','C','G','T'].inject(0.0) { |sum,l| sum += @matrix[l][i]**2 * bckgr[l] } - ['A','C','G','T'].inject(0.0) { |sum,l| sum += matrix[l][i] * bckgr[l] }**2
+        m2 += deltai
+      }
+    end
+    
+    def p_value(threshold, mean = nil, variance = nil)
+      mean = mean ? mean : score_mean
+      variance = variance ? variance : score_variance
+      n_ = (threshold - mean) / Math.sqrt(variance)
+      p_value = (1 - Math.erf2(n_/Math.sqrt(2))) / 2.0
+    end
+    
+    def best_word
+      return (0...size).inject("") { |word, i|
+        max = ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.max
+        maxlets = ['A', 'C', 'G', 'T'].select { |l| @matrix[l][i] == max }
+        word << (maxlets.size == 1 ? maxlets.first : "N")
+      }
+    end
+    
+    def strict_consensus
+      return IUPAC.new((0...size).inject("") { |word, i|
+        max = ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.max
+        maxlets = ['A', 'C', 'G', 'T'].inject("") { |lets, l| lets += @matrix[l][i] == max ? l : ""}
+        word += IUPAC::CODE[maxlets]
+      })
+    end
+    
+    def consensus_string(beautiful = false)
+      checkerr("words count is undefined") { !@words_count }
+      i2o4, thc, tlc = icd2of4, icdThc, icdTlc
+      icd = infocod
+      
+      return String.new((0...size).inject("") { |word, i|
+
+        scores = ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.uniq.sort.reverse
+        
+        if icd[i] > i2o4
+          scores = [scores.first]
+        elsif icd[i] > thc
+          scores = scores[0..1]
+        elsif icd[i] > tlc
+          scores = scores[0..2]
+        end
+        
+        lets = ['A', 'C', 'G', 'T'].inject("") { |lets, l| lets += scores.include?(@matrix[l][i]) ? l : ""}
+        
+        reslet = IUPAC::CODE[lets]
+        reslet = reslet.downcase if beautiful && lets.size > 2
+        
+        word += reslet
+      })
+    end
+    
+    def consensus
+      checkerr("words count is undefined") { !@words_count }
+      i2o4, thc, tlc = icd2of4, icdThc, icdTlc
+      icd = infocod
+      
+      return IUPAC.new((0...size).inject("") { |word, i|
+
+        scores = ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.uniq.sort.reverse
+        
+        if icd[i] > i2o4
+          scores = [scores.first]
+        elsif icd[i] > thc
+          scores = scores[0..1]
+        elsif icd[i] > tlc
+          scores = scores[0..2]
+        end
+        
+        lets = ['A', 'C', 'G', 'T'].inject("") { |lets, l| lets += scores.include?(@matrix[l][i]) ? l : ""}
+        
+        word += IUPAC::CODE[lets]
+      })
+    end
+    
+    def find_hit(s, score_g, use2strands = true)
+      (0..(s.size - @size)).each { |i|
+        seq, seq_rc = s[i, @size], s[i, @size].revcomp!
+        score_p, score_rc = score(seq), score(seq_rc)
+        r = use2strands ? [score_p,score_rc].max : score_p
+        return i if r >= score_g
+      }
+      return nil
+    end
+    
+    def find_hits(s, score_g, use2strands = true)
+      (0..(s.size - @size)).select { |i|
+        seq, seq_rc = s[i, @size], s[i, @size].revcomp!
+        score_p, score_rc = score(seq), score(seq_rc)
+        r = use2strands ? [score_p,score_rc].max : score_p
+        r >= score_g ? i : nil
+      }.compact
+    end
+    
+    def collect_hits(s, score_g, use2strands = true)
+      result = []
+      (0..(s.size - @size)).each { |i|
+        seq, seq_rc = s[i, @size], s[i, @size].revcomp!
+        score_p, score_rc = score(seq.upcase), score(seq_rc.upcase)
+        result << [score_p, seq, false, i] if score_p >= score_g
+        result << [score_rc, seq_rc, true, i] if score_rc >= score_g
+      }
+      result
+    end
+    
+    def best_hit(s, use2strands = true)
+      
+      checkerr("too short sequence") { s.size < @size }
+      return (0..(s.size - @size)).inject(-Float::MAX) { |r, i|
+        seq, seq_rc = s[i, @size], s[i, @size].revcomp!
+        score_p, score_rc = score(seq), score(seq_rc)
+        r = use2strands ? [r,score_p,score_rc].max : [r,score_p].max
+      }
+    end
+    
+    def eql?(pm)
+      return ['A','C','G','T'].inject(true) { |equal, letter|
+        equal = equal && @matrix[letter].eql?(pm.matrix[letter])
+      }
+    end
+    
+    def flexeql?(pm)
+      checkerr("for what?") { true }
+      return ['A','C','G','T'].inject(true) { |equal, letter|
+        # report "letter=#{letter}"
+        equal = equal && (0...@size).inject(true) { |deepequal, position| 
+          # report "position=#{position}, delta=#{@matrix[letter][position] - pm.matrix[letter][position]}"
+          deepequal = deepequal && (@matrix[letter][position] - pm.matrix[letter][position]).abs < 10**-11 
+        }
+      }
+    end
+    
+    def initialize(size, matrix = nil, words_count = nil)
+      checkerr("matrix['A'].size != size, #{matrix['A'].size} != #{size}") { matrix != nil && size != matrix['A'].size }
+      @size = size
+      @matrix = matrix == nil ? PM.new_matrix(size) : matrix
+      if !words_count || words_count <= 0
+        words_count = col_sum(0)
+        @words_count = words_count.round >= 2 ? words_count.round : nil
+      else
+        @words_count = words_count
+      end
+    end
+    
+    def col_sum(index = 0, letset = ['A','C','G','T'])
+      return letset.inject(0) { |sum, l| sum += @matrix[l][index] }
+    end
+    
+    def PM.col_sum(matrix, index = 0)
+      return matrix['A'][index] + matrix['C'][index] + matrix['G'][index] + matrix['T'][index]
+    end
+      
+    def to_pwm!(words_count = nil, probs = Randoom::DEF_PROBS, pseudocount = 1)
+      @words_count = words_count if words_count && words_count > 0
+      
+      @matrix.each_key do |letter|
+        (0...@size).each { |pos|
+          
+          #p "pcm"
+          #p @matrix[letter][pos]
+          #p @matrix[letter][pos] + (probs[letter] * pseudocount)
+          #p ( (@words_count + pseudocount) * probs[letter])
+          #exit
+          
+          @matrix[letter][pos] = Math::log( (@matrix[letter][pos] + (probs[letter] * pseudocount)) / ( (@words_count + pseudocount) * probs[letter]) )
+          
+        }
+      end
+      
+      return self
+    end
+    
+    def get_pwm(words_count = nil, probs = Randoom::DEF_PROBS, pseudocount = 1)
+      return self.dup.to_pwm!(words_count, probs, pseudocount)
+    end
+    alias to_pwm get_pwm
+    
+    def get_ppm(words_count = nil)
+      words_count = @words_count unless words_count
+      checkerr("undefined words count") { !words_count || words_count <= 0 }
+      ppm = @matrix['N'] ?  PM.new_matrix_iupac(@size) : PM.new_matrix(@size)
+      @matrix.each_key { |letter|
+        (0...@size).each { |i|
+          ppm[letter][i] = @matrix[letter][i].to_f / words_count
+        }
+      }
+      return PPM.new(@size, ppm, words_count)
+    end
+    alias to_ppm get_ppm
+    
+    def score(word)
+      checkerr("word size != pwm.size") { @size != word.size }
+      checkerr("word #{word} has strange characters") { 
+        @matrix.keys.include?('N') ? word.tr('ACGTRYKMSWBDHVN', '').size > 0 : word.tr('ACGT', '').size > 0
+      }
+      return (0...@size).inject(0) { |sum, i| 
+        sum += @matrix[word[i,1]][i]
+      }
+    end
+    
+    def best_score
+      return (0...size).inject(0) { |sum, i|
+        sum += ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.max
+      }
+    end
+    
+    def worst_score
+      return (0...size).inject(0) { |sum, i|
+        sum += ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.min
+      }
+    end
+    
+    def dup
+      new_matrix = {}
+      @matrix.each_key { |letter| new_matrix[letter] = @matrix[letter].dup }
+      return PM.new(@size, new_matrix, @words_count)
+    end
+    
+    def PM.new_pcm(words, iupacomp = false)
+      size = words[0].size
+      counts = PM.new_matrix(size)
+      counts.each_value { |arr| arr.fill(0) }
+      words.each { |word|
+        0.upto(size-1) { |i|
+          letter = word[i,1].upcase
+          checkerr("unknown letter #{letter}") { !['A', 'C', 'G', 'T', 'N'].include?(letter) }
+          if letter != 'N'
+            counts[letter][i] += 1
+          else
+            ['A', 'C', 'G', 'T'].each { |l| counts[l][i] += 0.25 }
+          end
+        }
+      }
+      newpcm = PM.new(size, counts, words.size)
+      newpcm.iupacomp! if iupacomp
+      return newpcm
+    end
+    
+    def PM.new_pwm(words)
+      pcm = PM.new_pcm(words)
+      pcm.to_pwm!
+      return pcm
+    end
+    
+    def PM.load(filename)
+      # supporting pat & pwm formats (letter-column and letter-row format)
+      input = IO.read(filename)
+      tm = []
+      input.each_line { |line| 
+        l_a = line.split
+        begin
+          l_a = l_a.collect { |a_i| Float(a_i) }
+        rescue
+          next
+        end
+        tm << l_a
+      }
+      tm = tm.transpose if tm.size == 4
+      matrix = PM.new_matrix(tm.size)
+      tm.each_index { |i| ['A', 'C', 'G', 'T'].each_with_index { |l, j| matrix[l][i] = tm[i][j] }  }
+      
+      ppm_mode = (0...tm.size).inject(true) { |ppm_ya, i| ppm_ya &= col_sum(matrix, i).round == 1 }
+      
+      return ppm_mode ? PPM.new(tm.size, matrix) : PM.new(tm.size, matrix)
+    end
+    
+    def save(filename)
+      File.open(filename, "w") { |out_f|
+        case File.ext_wo_name(filename)
+        when "pwm"
+          ['A', 'C', 'G', 'T'].each { |letter|
+            @matrix[letter].each { |e|
+              out_f << "#{e} "
+            }
+            out_f << $/
+          }
+        when "pat"
+          out_f.puts File.name_wo_ext(filename)
+          (0...@size).each { |i|
+            ['A', 'C', 'G', 'T'].each { |letter|
+              out_f << "#{@matrix[letter][i]} "
+            }
+            out_f << $/
+          }
+        when "xml"
+          checkerr("small-BiSMark is not supported at this moment")
+        else
+          checkerr("unknown motif file format specified")
+        end
+      }
+    end
+    
+    def positiv!
+      min = @matrix.values.collect { |v| v.min }.min.abs
+      @matrix.each_value { |v| (0...v.size).each { |i| v[i] += min } }
+      return self
+    end
+    
+    def revcomp!
+      @matrix['A'], @matrix['T'] = @matrix['T'], @matrix['A']
+      @matrix['C'], @matrix['G'] = @matrix['G'], @matrix['C']
+      @matrix.each_value { |v| v.reverse! }
+      self
+    end
+    
+    def to_bismark(b)
+      pwm = @matrix['A'][0].is_a?(Float)
+      attributes = {"length" => @size}
+      attributes["words-count"] = @words_count if @words_count && @words_count > 0
+      pe = b.add_element( pwm ? "PWM" : "PCM", attributes )
+      (0...@matrix['A'].size).each { |i|
+        pm_c = pe.add_element("pm-column", {"position" => i+1})
+        ['A', 'C', 'G', 'T'].each { |l|
+          pm_c.add_element(l.downcase).add_text(@matrix[l][i].to_s)
+        }
+      }
+    end
+    
+    def PM.from_bismark(b, iupacomp = false)
+      
+      checkerr("empty small-BiSMark file?") { !b }
+      float_m = (b.name == "PPM" || b.name == "PWM" || b.name == "WPCM")
+      words_count = b.attributes["words-count"] ? b.attributes["words-count"].to_f : nil
+      
+      matrix = {"A" => [], "C" => [], "G" => [], "T" => []}
+      b.elements.each("pm-column") { |pmc|
+        position = pmc.attributes["position"].to_i
+        ['A', 'C', 'G', 'T'].each { |l| matrix[l][position-1] = float_m ? pmc.elements[l.downcase].get_text.to_s.to_f : pmc.elements[l.downcase].get_text.to_s.to_i }
+      }
+      if b.name == "PPM"
+        newppm = PPM.new(matrix['A'].size, matrix, words_count)
+        newppm.iupacomp! if iupacomp
+        return newppm
+      end
+      if b.name == "PCM"
+        @words_count = col_sum(matrix)
+        newpcm = PM.new(matrix['A'].size, matrix, words_count)
+        newpcm.iupacomp! if iupacomp
+        return newpcm
+      end
+      if b.name == "PWM" && iupacomp
+        raise "cannot force IUPAC compatible PWM"
+      end
+      return PM.new(matrix['A'].size, matrix, words_count)
+    end
+    
+    IUPAC_LS = (IUPAC::CODE.keys - ['A','C','G','T']).collect { |iul| [IUPAC::CODE[iul], iul.split(//)] }
+    def iupacomp!
+      @words_count = (0...@size).collect { |i| col_sum(i) }.max unless @words_count # for unbalanced matrices (Genomatix has some)
+      # @words_count = @words_count.round < 2.0 ? nil : @words_count.round
+      
+      IUPAC_LS.each { |iul_ls|
+        @matrix[iul_ls[0]] = (0...@size).collect { |i| col_sum(i, iul_ls[1]) / iul_ls[1].size }
+      }
+      
+      return self
+    end
+    
+    def m3sd(bckgr = Randoom::DEF_PROBS)
+    
+      mean = (0...@size).inject(0.0) { |mean, i| mean += ['A','C','G','T'].inject(0.0) { |sum,l| sum += @matrix[l][i] * bckgr[l] } }
+      dev = (0...@size).inject(0.0) { |m2, i| 
+        deltai = ['A','C','G','T'].inject(0.0) { |sum,l| sum += @matrix[l][i]**2 * bckgr[l] } - ['A','C','G','T'].inject(0.0) { |sum,l| sum += matrix[l][i] * bckgr[l] }**2
+        m2 += deltai
+      }
+      sigma = Math.sqrt(dev)
+      
+      mean+3*sigma
+    end
+    
+    def fixwc
+      return unless @words_count
+      @words_count = (0...@size).collect { |i| col_sum(i) }.max
+    end
+    
+    protected
+    def PM.new_matrix(size)
+      return {
+        'A' => Array.new(size),
+        'C' => Array.new(size),
+        'G' => Array.new(size),
+        'T' => Array.new(size) }
+    end
+    
+    def PM.new_matrix_iupac(size)
+      return {
+        'A' => Array.new(size),
+        'C' => Array.new(size),
+        'G' => Array.new(size),
+        'T' => Array.new(size),
+        'R' => Array.new(size),
+        'Y' => Array.new(size),
+        'K' => Array.new(size),
+        'M' => Array.new(size),
+        'S' => Array.new(size),
+        'W' => Array.new(size),
+        'B' => Array.new(size),
+        'D' => Array.new(size),
+        'H' => Array.new(size),
+        'V' => Array.new(size),
+        'N' => Array.new(size)
+        }
+    end
+    
+  end
+
+  class PPM < PM
+    
+    #DEPRECATED, use iupacomp! instead
+    #def make_N_comp!
+    #  @matrix['N'] = (0...size).collect { 0.25 }
+    #  return self
+    #end
+    
+    def initialize(size, matrix = nil, words_count = nil)
+      checkerr("matrix['A'].size != size") { matrix != nil && size != matrix['A'].size }
+      @size = size
+      @matrix = matrix == nil ? PM.new_matrix(size) : matrix
+      @words_count = words_count
+    end
+    
+    def iupacomp!
+      @words_count = 4.0 unless @words_count
+      
+      IUPAC_LS.each { |iul_ls|
+        @matrix[iul_ls[0]] = (0...@size).collect { |i| col_sum(i, iul_ls[1]) / iul_ls[1].size }
+      }
+      
+      return self
+    end
+    
+    def score(word)
+      checkerr("word size != ppm.size") { @size != word.size }
+      checkerr("word #{word} has strange characters") { 
+        @matrix.keys.include?('N') ? word.tr('ACGTRYKMSWBDHVN', '').size > 0 : word.tr('ACGT', '').size > 0
+      }
+      return (0...@size).inject(1) { |mul, i| 
+        mul *= @matrix[word[i,1]][i]
+      }
+    end
+    
+    def best_score
+      return (0...size).inject(1) { |mul, i|
+        mul *= ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.max
+      }    
+    end
+    
+    def worst_score
+      return (0...size).inject(0) { |mul, i|
+        mul *= ['A', 'C', 'G', 'T'].collect { |l| @matrix[l][i] }.min
+      }
+    end
+    
+    def to_bismark(b)
+      attributes = {"length" => @size}
+      attributes["words-count"] = @words_count if @words_count
+      pe = b.add_element("PPM", attributes)
+      (0...@matrix['A'].size).each { |i|
+        pm_c = pe.add_element("pm-column", {"position" => i+1})
+        ['A', 'C', 'G', 'T'].each { |l|
+          pm_c.add_element(l.downcase).add_text(@matrix[l][i].to_s)
+        }
+      }
+    end
+    
+    def PPM.probs2IUPAC!(probs)
+      IUPAC_LS.each { |iul_ls|
+        probs[iul_ls[0]] = iul_ls[1].inject(0) { |sum, l| sum += probs[l] } / iul_ls[1].size
+      }
+      return probs
+    end
+    
+    def get_pwm(words_count = nil, probs = Randoom::DEF_PROBS, pseudocount = 1.0)
+      
+      probs = PPM.probs2IUPAC!(probs.dup)
+      
+      words_count = @words_count if !words_count || words_count == 0
+      checkerr("undefined words count") { !words_count }
+      
+      pwm = @matrix['N'] ? PM.new_matrix_iupac(@size) : PM.new_matrix(@size)
+      
+      @matrix.each_key do |letter|
+        (0...@size).each { |pos|
+          
+          pwm[letter][pos] = Math::log( (@matrix[letter][pos] * words_count + (probs[letter] * pseudocount) ) / ( (words_count + pseudocount) * probs[letter]) )
+          
+        }
+      end
+      return PM.new(@size, pwm, words_count)
+      #pcm = get_pcm(words_count)
+      #pcm.iupacomp! if @matrix['N']
+      #return pcm.to_pwm!(words_count, probs, pseudocount)
+    end
+    alias to_pwm get_pwm
+    
+    def get_pwm0pc(probs = Randoom::DEF_PROBS)
+      new_matrix = {}
+      @matrix.each_key { |letter| new_matrix[letter] = @matrix[letter].dup }
+      newpm = PM.new(@size, new_matrix, nil)
+      
+      new_matrix.each_key do |letter|
+        (0...@size).each { |pos|
+          new_matrix[letter][pos] = Math::log(@matrix[letter][pos] / probs[letter])
+        }
+      end
+      
+      return newpm
+    end
+    
+    def to_pwm!
+      raise "cannot force PPM class to PWM, use to_pwm instead"
+    end
+    
+    def get_pcm(words_count = nil)
+      words_count = @words_count unless words_count
+      checkerr("undefined words count") { !words_count }
+      counts = PM.new_matrix(@size)
+      (0...size).each { |i|
+        ['A', 'C', 'G', 'T'].each { |l|
+          counts[l][i] = @matrix[l][i] * words_count
+        }
+      }
+      newpcm = PM.new(size, counts, words_count).iupacomp!
+      return newpcm
+    end
+    alias to_pcm get_pcm
+    
+    def PPM.from_IUPAC(iupac)
+      matrix = {"A" => [], "C" => [], "G" => [], "T" => []}
+      
+      (0...iupac.size).each { |i|
+        matrix.each_key { |k| matrix[k] << 0.0 }
+        letters = IUPAC::REVCODE[iupac[i]]
+        (0...letters.size).each { |j|
+          matrix[letters[j]][-1] = 1.0/letters.size
+        }
+      }
+      
+      newppm = PPM.new(iupac.size, matrix, 4.0)
+      newppm.iupacomp!
+      
+      newppm
+    end
+    
+  end
+end