forester/ruby/evoruby/lib/evo/tool/fasta_extractor.rb

   1 #
   2 # = lib/evo/apps/fasta_extractor.rb - FastaExtractor class
   3 #
   4 # Copyright::  Copyright (C) 2006-2008 Christian M. Zmasek
   5 # License::    GNU Lesser General Public License (LGPL)
   6 #
   7 # $Id: fasta_extractor.rb,v 1.2 2010/12/13 19:00:11 cmzmasek Exp $
   8
   9
  10 require 'lib/evo/util/util'
  11 require 'lib/evo/util/constants'
  12 require 'lib/evo/util/command_line_arguments'
  13
  14
  15 module Evoruby
  16
  17     class FastaExtractor
  18
  19         PRG_NAME                           = "fae"
  20         PRG_VERSION                        = "1.0.0"
  21         PRG_DESC                           = "extraction of nucleotide sequences from a fasta file by names from wublast search"
  22         PRG_DATE                           = "2008.08.09"
  23         COPYRIGHT                          = "2008-2009 Christian M Zmasek"
  24         CONTACT                            = "phylosoft@gmail.com"
  25         WWW                                = "www.phylosoft.org"
  26         HELP_OPTION_1                      = 'help'
  27         HELP_OPTION_2                      = 'h'
  28
  29
  30         def run()
  31
  32             Util.print_program_information( PRG_NAME,
  33                 PRG_VERSION,
  34                 PRG_DESC ,
  35                 PRG_DATE,
  36                 COPYRIGHT,
  37                 CONTACT,
  38                 WWW,
  39                 STDOUT )
  40
  41             ld = Constants::LINE_DELIMITER
  42
  43             begin
  44                 cla = CommandLineArguments.new( ARGV )
  45             rescue ArgumentError => e
  46                 Util.fatal_error( PRG_NAME, "error: " + e.to_s )
  47             end
  48
  49             if ( cla.is_option_set?( HELP_OPTION_1 ) ||
  50                      cla.is_option_set?( HELP_OPTION_2 ) )
  51                 print_help
  52                 exit( 0 )
  53             end
  54
  55             if ( cla.get_number_of_files != 3 )
  56                 print_help
  57                 exit( -1 )
  58             end
  59
  60             allowed_opts = Array.new
  61
  62             disallowed = cla.validate_allowed_options_as_str( allowed_opts )
  63             if ( disallowed.length > 0 )
  64                 Util.fatal_error( PRG_NAME,
  65                     "unknown option(s): " + disallowed,
  66                     STDOUT )
  67             end
  68
  69             input_file  = cla.get_file_name( 0 )
  70             names_file  = cla.get_file_name( 1 )
  71             output_file = cla.get_file_name( 2 )
  72
  73             if  !File.exist?( input_file )
  74                 Util.fatal_error( PRG_NAME, "error: input file [#{input_file}] does not exist" )
  75             end
  76             if  !File.exist?( names_file )
  77                 Util.fatal_error( PRG_NAME, "error: names file [#{names_file}] does not exist" )
  78             end
  79             if File.exist?( output_file   )
  80                 Util.fatal_error( PRG_NAME, "error: [#{output_file }] already exists" )
  81             end
  82
  83             names = extract_names_with_frames( names_file )
  84
  85             extract_sequences( names, input_file, output_file )
  86
  87             puts
  88             Util.print_message( PRG_NAME, "OK" )
  89             puts
  90
  91         end
  92
  93
  94         def extract_names_with_frames( names_file )
  95             names = Hash.new()
  96             File.open( names_file ) do | file |
  97                 while line = file.gets
  98                     if ( !Util.is_string_empty?( line ) && !(line =~ /\s*#/ ) )
  99                         if ( line =~ /(\S+)\s+([+|-]\d)\s+\d+\s+(\S+)/ )
 100                             name  = $1
 101                             frame = $2
 102                             e     = $3
 103                             names[ name ] =  "[" + frame + "] [" + e + "]"
 104                         end
 105                     end
 106                 end
 107             end
 108             names
 109         end
 110
 111         def extract_sequences( names, fasta_file, output_file )
 112             output = File.open( output_file, "a" )
 113             matching_state = false
 114             counter = 0
 115             File.open( fasta_file ) do | file |
 116                 while line = file.gets
 117                     if !Util.is_string_empty?( line )
 118                         if ( line =~ /\s*>\s*(.+)/ )
 119                             name = $1
 120                             if names.has_key?( name )
 121                                 matching_state = true
 122                                 counter += 1
 123                                 puts counter.to_s + ". " +name + " " + names[ name ]
 124                                 output.print( ">" + name + " " + names[ name ] )
 125                                 output.print( Evoruby::Constants::LINE_DELIMITER )
 126                             else
 127                                 matching_state = false
 128                             end
 129                         elsif matching_state
 130                             output.print( line )
 131                         end
 132                     end
 133                 end
 134             end
 135             output.close()
 136         end
 137
 138         def print_help()
 139             puts( "Usage:" )
 140             puts()
 141             puts( "  " + PRG_NAME + ".rb <input fasta file> <names file based on blast output> <output file>" )
 142             puts()
 143         end
 144
 145     end # class FastaExtractor
 146 end