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

   1 #
   2 # = lib/evo/tool/multi_domain_seq_extractor - MultiDomainSeqExtractor class
   3 #
   4 # Copyright::    Copyright (C) 2017 Christian M. Zmasek
   5 # License::      GNU Lesser General Public License (LGPL)
   6
   7 require 'lib/evo/util/constants'
   8 require 'lib/evo/util/util'
   9 require 'lib/evo/util/command_line_arguments'
  10 require 'lib/evo/io/parser/hmmscan_multi_domain_extractor'
  11
  12 module Evoruby
  13   class MultiDomainSeqExtractor
  14
  15     PRG_NAME       = "mdsx"
  16     PRG_VERSION    = "1.000"
  17     PRG_DESC       = "Extraction of multi domain sequences from hmmscan output"
  18     PRG_DATE       = "20170220"
  19     WWW            = "https://sites.google.com/site/cmzmasek/home/software/forester"
  20
  21     LOG_FILE_SUFFIX                    = '_MDSX.log'
  22     HELP_OPTION_1                      = 'help'
  23     HELP_OPTION_2                      = 'h'
  24     def run()
  25
  26       Util.print_program_information( PRG_NAME,
  27       PRG_VERSION,
  28       PRG_DESC ,
  29       PRG_DATE,
  30       WWW,
  31       STDOUT )
  32
  33       if ( ARGV == nil || ( ARGV.length < 1 )  )
  34         print_help
  35         exit( -1 )
  36       end
  37
  38       begin
  39         cla = CommandLineArguments.new( ARGV )
  40       rescue ArgumentError
  41         Util.fatal_error( PRG_NAME, "error: " + $!, STDOUT )
  42       end
  43
  44       if ( cla.is_option_set?( HELP_OPTION_1 ) ||
  45       cla.is_option_set?( HELP_OPTION_2 ) )
  46         print_help
  47         exit( 0 )
  48       end
  49
  50       unless ( cla.get_number_of_files == 2 || cla.get_number_of_files == 3 )
  51         print_help
  52         exit( -1 )
  53       end
  54
  55       allowed_opts = Array.new
  56
  57       disallowed = cla.validate_allowed_options_as_str( allowed_opts )
  58       if ( disallowed.length > 0 )
  59         Util.fatal_error( PRG_NAME,
  60         "unknown option(s): " + disallowed,
  61         STDOUT )
  62       end
  63
  64       domain_id           = cla.get_file_name( 0 )
  65       hmmscan_output      = cla.get_file_name( 1 )
  66       fasta_sequence_file = ""
  67       outfile             = ""
  68
  69       if (cla.get_number_of_files == 3)
  70         fasta_sequence_file = cla.get_file_name( 2 )
  71       else
  72         hmmscan_index = hmmscan_output.index(Constants::HMMSCAN)
  73         if ( hmmscan_index != nil )
  74           prefix = hmmscan_output[0 .. hmmscan_index-1 ]
  75           suffix = Constants::ID_NORMALIZED_FASTA_FILE_SUFFIX
  76           files = Dir.entries( "." )
  77           matching_files = Util.get_matching_files( files, prefix, suffix)
  78           if matching_files.length < 1
  79             Util.fatal_error( PRG_NAME, 'no file matching [' + prefix +
  80             '...' + suffix + '] present in current directory: need to indicate <file containing complete sequences in fasta format> as second argument' )
  81           end
  82           if matching_files.length > 1
  83             Util.fatal_error( PRG_NAME, 'more than one file matching [' +
  84             prefix  + '...' + suffix + '] present in current directory: need to indicate <file containing complete sequences in fasta format> as second argument' )
  85           end
  86           fasta_sequence_file = matching_files[ 0 ]
  87         else
  88           Util.fatal_error( PRG_NAME, 'input files do not seem in format for standard analysis pipeline, need to explicitly indicate all' )
  89         end
  90       end
  91       hmmscan_index = hmmscan_output.index(Constants::HMMSCAN)
  92       if ( hmmscan_index != nil )
  93         outfile = hmmscan_output.sub(Constants::HMMSCAN, "_") + "_MDSX"
  94       else
  95         Util.fatal_error( PRG_NAME, 'input files do not seem in format for standard analysis pipeline, need to explicitly indicate all' )
  96       end
  97
  98       log = String.new
  99       ld = Constants::LINE_DELIMITER
 100
 101       #      puts()
 102       #      puts( "Domain                                                                             : " + domain_id )
 103       #      log << "Domain                                                                             : " + domain_id + ld
 104       #      puts( "Hmmscan outputfile                                                                 : " + hmmscan_output )
 105       #      log << "Hmmscan outputfile                                                                 : " + hmmscan_output + ld
 106       #      puts( "Fasta sequencefile (complete sequences)                                            : " + fasta_sequence_file )
 107       #      log << "Fasta sequencefile (complete sequences)                                            : " + fasta_sequence_file + ld
 108       #      puts( "Outputfile                                                                         : " + outfile + ".fasta" )
 109       #      log << "Outputfile                                                                         : " + outfile + ld
 110       #      puts( "Passed sequences outfile (fasta)                                                   : " + outfile + PASSED_SEQS_SUFFIX )
 111       #      log << "Passed sequences outfile (fasta)                                                   : " + outfile + PASSED_SEQS_SUFFIX + ld
 112       #      puts( "Failed sequences outfile (fasta)                                                   : " + outfile + FAILED_SEQS_SUFFIX )
 113       #      log << "Failed sequences outfile (fasta)                                                   : " + outfile + FAILED_SEQS_SUFFIX + ld
 114       #      puts( "Logfile                                                                            : " + outfile + LOG_FILE_SUFFIX )
 115       #      log << "Logfile                                                                            : " + outfile + LOG_FILE_SUFFIX + ld
 116
 117       puts
 118       log <<  ld
 119
 120       domain_count = 0
 121       begin
 122         parser = HmmscanMultiDomainExtractor.new()
 123         domain_count = parser.parse( domain_id,
 124         hmmscan_output,
 125         fasta_sequence_file,
 126         outfile,
 127         log )
 128       rescue ArgumentError, IOError => e
 129         Util.fatal_error( PRG_NAME, "error: " + e.to_s, STDOUT )
 130
 131       rescue Exception => e
 132         puts e.backtrace
 133         Util.fatal_error( PRG_NAME, "unexpected exception: " + e.to_s, STDOUT )
 134
 135       end
 136
 137       puts
 138       Util.print_message( PRG_NAME, "extracted a total of " + domain_count.to_s + " domains" )
 139       #  Util.print_message( PRG_NAME, "wrote: " + outfile + ".fasta")
 140       #  Util.print_message( PRG_NAME, "wrote: " + outfile + LOG_FILE_SUFFIX )
 141       #  Util.print_message( PRG_NAME, "wrote: " + outfile + PASSED_SEQS_SUFFIX )
 142       #  Util.print_message( PRG_NAME, "wrote: " + outfile + FAILED_SEQS_SUFFIX )
 143
 144       begin
 145         f = File.open( outfile + LOG_FILE_SUFFIX, 'a' )
 146         f.print( log )
 147         f.close
 148       rescue Exception => e
 149         Util.fatal_error( PRG_NAME, "error: " + e.to_s )
 150       end
 151
 152       puts
 153       Util.print_message( PRG_NAME, "OK" )
 154       puts
 155
 156     end
 157
 158     def print_help()
 159       puts()
 160       puts( "Usage:" )
 161       puts()
 162       puts( "  " + PRG_NAME + ".rb <da> <hmmscan outputfile> [file containing complete sequences in fasta format]" )
 163       puts()
 164       puts( "  options: -"  )
 165       puts()
 166       puts( "Examples:" )
 167       puts
 168       puts( "  " + PRG_NAME + ".rb " )
 169       puts
 170
 171       puts()
 172     end
 173
 174   end # class DomainSequenceExtractor
 175
 176 end # module Evoruby