class BuildSplitsets @@help = [] @@help << "build_splitset.rb [options]" @@help << "Options:" @@help << "-i \t inputfile with sequences" @@help << "-o \t path where to put the generated split sets training/testing(crossvalidation) and hold-out (if not existent it will be generated)" @@help << "-f \t x fold crossvalidation" @@help << "-s \t how much percent of the sequences shoud be used for the training/testing set. The remaining sequences are used for the hold-out set" @@help << "-p \t plot the distributions of the GO-Numbers" @@help << "-h(elp) print this help" def initialize(opts) @xfold = 0 @split = 0 @seqs = [] @seqid = 0 @plotfile = "" @plot =false @outpath = "" @cv_files = [] i = 0 # parse options while i < opts.size if opts[i].eql?("-f") if opts[i+1] =~ /\d+/ @xfold = opts[i+1].to_i i+=1 else raise "Option -f needs an integer parameter!" end elsif opts[i].eql?("-s") if opts[i+1] =~ /\d+/ @split = opts[i+1].to_i i+=1 else raise "Option -s needs an integer parameter!" end elsif opts[i].eql?("-i") if File.exists?(opts[i+1]) file = File.open(opts[i+1],'rb') lines = file.readlines parseSequences(lines) if opts[i+1] =~ /(\d+)\.f/ # for the naming of the plot files @seqid = $1 end i+=1 else raise "File #{opts[i+1]} does not exist!" end elsif opts[i].eql?("-p") @plot = true elsif opts[i].eql?("-o") if !(opts[i+1].length.nil?) @outpath = File.expand_path(opts[i+1])+"/" i+=1 else raise "-o needs a directoryname" end elsif opts[i].eql?("-h") || opts[i].eql?("-help") puts @@help exit(0) else raise "Unknown option #{opts[i]}" end i+=1 end if @outpath.eql?("") puts "ERROR: option -o has to be set!" puts @@help exit(0) elsif @xfold == 0 puts "ERROR: option -f has to be set and to be greater than 0!" puts @@help exit(0) elsif @split == 0 raise "ERROR: option -s has to be set and to be greater than 0!" puts @@help exit(0) elsif @seqs.size < 1 raise "ERROR: option -i has to be set!" puts @@help exit(0) end #@outpath = @outpath+@seqid+"/" # uncomment for xargs command system "mkdir #{@outpath}" #get go stats and make tabular file for plotting if -p option is set if @plot makePlots end #randomize sequences random = @seqs.sort_by{rand(@seqs.size)} @seqs = random #split in test/training and hold-out set buildSplitsets #build combinatoric file for crossvalidation buildCombinatoricFile end def parseSequences(lines) i = 0 s = "" while i < lines.size if lines[i] =~ /^>/ s = lines[i] i+=1 while !(lines[i] =~ /^>/) && !(lines[i].nil?) s+=lines[i] i+=1 end @seqs << s s = "" else raise lines[i] end end end def makePlots #first plot, x-GO:Number, y-frequency go_counter = Hash.new z = 0 @seqs.each do |s| if s =~ /((GO:\d+,{0,1})+)/ z = z+1 gos = $1 go_numbers = gos.split(",") go_numbers.each do |g| if go_counter[g].nil? go_counter[g] = 1 else go_counter[g]+=1 end end else raise s end end plotfile = @outpath+"plotfile_1" p_out = File.open(plotfile,'wb') go_counter.each_key do |k| p_out.write k+"\t"+go_counter[k].to_s+"\n" end p_out.close r_file = File.open(plotfile+".R", 'w') r_file.write("x <- read.table(\"#{plotfile}\", header=F)\n") r_file.write("plot(x[[1]],x[[2]], main=\"GO-Number frequency\", xlab=\"GO-Number\", ylab=\"frequency\", type=\"l\")") r_file.close system "R -f #{plotfile}.R " system "mv ./Rplots.pdf #{@outpath}Rplots#{@seqid}_1.pdf" #second plot, x-frequency, y-amount of x-frequency occurred frequency_counter = Hash.new go_counter.each_key do |g| if frequency_counter[go_counter[g].to_s].nil? frequency_counter[go_counter[g].to_s] = 1 else frequency_counter[go_counter[g].to_s]+=1 end end a = [] i = 0 frequency_counter.each_key do |k| a[k.to_i] = frequency_counter[k] end while i < a.size if a[i].nil? a[i] = 0 # elsif i < 10 #set the threshold here to remove overrepresentatives in the plot # a[i] = 0 end i+=1 end plotfile = @outpath+"plotfile_2" p_out = File.open(plotfile,'wb') a.each_index do |k| p_out.write k.to_s+"\t"+a[k].to_s+"\n" end p_out.close r_file = File.open(plotfile+".R", 'w') r_file.write("x <- read.table(\"#{plotfile}\", header=F)\n") r_file.write("plot(x[[1]],x[[2]], main=\"GO-Number frequency distribution\", xlab=\"frequency\", ylab=\"occurrence\", type=\"l\")") r_file.close system "R -f #{plotfile}.R" system "mv ./Rplots.pdf #{@outpath}Rplots#{@seqid}_2.pdf" end def buildSplitsets puts "#{@seqs.size} sequences in sequence file!" n = (@seqs.size).to_f * (@split.to_f / 100.0) puts "There will be #{n} sequences within the training/testset!" n = n.to_i i = 0 tt_set = [] ho_set = [] while i < @seqs.size if i < n tt_set << @seqs[i] else ho_set << @seqs[i] end i+=1 end puts "The training/test set has #{tt_set.size} sequences!" puts "The hold-out set has #{ho_set.size} sequences!" #write hold-out set into jobfolder/holdoutXX_set.f holdout = File.open(@outpath+"holdout#{@seqid}_set.f", 'wb') ho_set.each {|seq| holdout.write(seq)} holdout.close #build the @xfold split set files n = (tt_set.size).to_f / (@xfold.to_f) puts "Each of the #{@xfold} crossvalidation sets will have #{n} sequences!" rest = tt_set.size % @xfold.to_f puts "#{rest} sequences will be equally distributed among the #{@xfold} sets!" n = n.to_i cross = [] i = 0 while i < @xfold j = 0 cross[i] = [] while j < n index = j+i*n cross[i] << tt_set[index] j+=1 end if rest > 0 cross[i] << tt_set[tt_set.size-1-rest] #its better to have balanced sets if the amount of sequences is not dividable by @xfold, rest = rest-1 #rather than pushing all of the remaining sequences in the last crossvalidation split set end i+=1 end #write all the crossvalidation splitsets in the jobfolder i = 0 cross.each do |split| i+=1 puts "Split #{i} has #{split.size} sequences!" cv_file = @outpath+"cv_split_#{@seqid}_#{i}.f" @cv_files << cv_file out = File.open(cv_file,'wb') split.each do |seq| out.write(seq) end out.close end end def buildCombinatoricFile combfile = File.open(@outpath+"crossvalidation_combinatorics.txt", 'w') @cv_files.each_index do |i| line = "" @cv_files.each_index do |j| if i != j # file i is added as the last one line = line+@cv_files[j]+"\t" end end line = line+@cv_files[i] combfile.write line+"\n" end combfile.close end end BuildSplitsets.new(ARGV)