forester/java/src/org/forester/application/gsdi.java

   1 // $Id:
   2 // FORESTER -- software libraries and applications
   3 // for evolutionary biology research and applications.
   4 //
   5 // Copyright (C) 2008-2009 Christian M. Zmasek
   6 // Copyright (C) 2008-2009 Burnham Institute for Medical Research
   7 // All rights reserved
   8 //
   9 // This library is free software; you can redistribute it and/or
  10 // modify it under the terms of the GNU Lesser General Public
  11 // License as published by the Free Software Foundation; either
  12 // version 2.1 of the License, or (at your option) any later version.
  13 //
  14 // This library is distributed in the hope that it will be useful,
  15 // but WITHOUT ANY WARRANTY; without even the implied warranty of
  16 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  17 // Lesser General Public License for more details.
  18 //
  19 // You should have received a copy of the GNU Lesser General Public
  20 // License along with this library; if not, write to the Free Software
  21 // Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
  22 //
  23 // Contact: phylosoft @ gmail . com
  24 // WWW: www.phylosoft.org/forester
  25
  26 package org.forester.application;
  27
  28 import java.io.File;
  29 import java.io.IOException;
  30 import java.io.Writer;
  31 import java.text.SimpleDateFormat;
  32 import java.util.ArrayList;
  33 import java.util.Date;
  34 import java.util.List;
  35
  36 import org.forester.io.parsers.PhylogenyParser;
  37 import org.forester.io.parsers.nhx.NHXParser;
  38 import org.forester.io.parsers.phyloxml.PhyloXmlParser;
  39 import org.forester.io.parsers.util.ParserUtils;
  40 import org.forester.io.writers.PhylogenyWriter;
  41 import org.forester.phylogeny.Phylogeny;
  42 import org.forester.phylogeny.PhylogenyMethods;
  43 import org.forester.phylogeny.PhylogenyNode;
  44 import org.forester.phylogeny.factories.ParserBasedPhylogenyFactory;
  45 import org.forester.phylogeny.factories.PhylogenyFactory;
  46 import org.forester.sdi.GSDI;
  47 import org.forester.sdi.SDI;
  48 import org.forester.sdi.SDI.TaxonomyComparisonBase;
  49 import org.forester.sdi.SDIse;
  50 import org.forester.sdi.SdiException;
  51 import org.forester.util.CommandLineArguments;
  52 import org.forester.util.ForesterConstants;
  53 import org.forester.util.ForesterUtil;
  54
  55 public final class gsdi {
  56
  57     private enum BASE_ALGORITHM {
  58         GSDI, SDI
  59     }
  60     final static public boolean REPLACE_UNDERSCORES_IN_NH_SPECIES_TREE     = true;
  61     final static private String STRIP_OPTION                               = "s";
  62     final static private String ALLOW_STRIPPING_OF_GENE_TREE_OPTION        = "g";
  63     final static private String SDI_OPTION                                 = "b";
  64     final static private String MOST_PARSIMONIOUS_OPTION                   = "m";
  65     final static private String GUESS_FORMAT_OF_SPECIES_TREE               = "q";
  66     final static private String HELP_OPTION_1                              = "help";
  67     final static private String HELP_OPTION_2                              = "h";
  68     final static private String DEFAULT_OUTFILE_SUFFIX                     = "_gsdi_out.phylo.xml";
  69     final static private String SUFFIX_FOR_LIST_OF_STIPPED_GENE_TREE_NODES = "_stripped_gene_tree_nodes.txt";
  70     final static private String LOGFILE_SUFFIX                             = "_gsdi_log.txt";
  71     final static private String PRG_NAME                                   = "gsdi";
  72     final static private String PRG_VERSION                                = "0.901";
  73     final static private String PRG_DATE                                   = "120608";
  74     final static private String PRG_DESC                                   = "general speciation duplication inference";
  75     final static private String E_MAIL                                     = "phylosoft@gmail.com";
  76     final static private String WWW                                        = "www.phylosoft.org/forester";
  77
  78     public static void main( final String args[] ) {
  79         try {
  80             ForesterUtil.printProgramInformation( PRG_NAME,
  81                                                   PRG_DESC,
  82                                                   PRG_VERSION,
  83                                                   PRG_DATE,
  84                                                   E_MAIL,
  85                                                   WWW,
  86                                                   ForesterUtil.getForesterLibraryInformation() );
  87             CommandLineArguments cla = null;
  88             try {
  89                 cla = new CommandLineArguments( args );
  90             }
  91             catch ( final Exception e ) {
  92                 ForesterUtil.fatalError( PRG_NAME, e.getMessage() );
  93             }
  94             if ( cla.isOptionSet( gsdi.HELP_OPTION_1 ) || cla.isOptionSet( gsdi.HELP_OPTION_2 ) ) {
  95                 System.out.println();
  96                 gsdi.print_help();
  97                 System.exit( 0 );
  98             }
  99             else if ( ( args.length < 2 ) || ( cla.getNumberOfNames() < 2 ) || ( cla.getNumberOfNames() > 3 ) ) {
 100                 System.out.println();
 101                 System.out.println( "Wrong number of arguments." );
 102                 System.out.println();
 103                 gsdi.print_help();
 104                 System.exit( -1 );
 105             }
 106             final List<String> allowed_options = new ArrayList<String>();
 107             allowed_options.add( gsdi.STRIP_OPTION );
 108             allowed_options.add( gsdi.SDI_OPTION );
 109             allowed_options.add( gsdi.GUESS_FORMAT_OF_SPECIES_TREE );
 110             allowed_options.add( gsdi.MOST_PARSIMONIOUS_OPTION );
 111             allowed_options.add( gsdi.ALLOW_STRIPPING_OF_GENE_TREE_OPTION );
 112             final String dissallowed_options = cla.validateAllowedOptionsAsString( allowed_options );
 113             if ( dissallowed_options.length() > 0 ) {
 114                 ForesterUtil.fatalError( gsdi.PRG_NAME, "unknown option(s): " + dissallowed_options );
 115             }
 116             execute( cla );
 117         }
 118         catch ( final IOException e ) {
 119             ForesterUtil.fatalError( gsdi.PRG_NAME, e.getMessage() );
 120         }
 121     }
 122
 123     private static void execute( final CommandLineArguments cla ) throws IOException {
 124         BASE_ALGORITHM base_algorithm = BASE_ALGORITHM.GSDI;
 125         boolean strip_species_tree = false;
 126         boolean most_parsimonous_duplication_model = false;
 127         boolean species_tree_in_phyloxml = true;
 128         boolean allow_stripping_of_gene_tree = false;
 129         if ( cla.isOptionSet( gsdi.STRIP_OPTION ) ) {
 130             strip_species_tree = true;
 131         }
 132         if ( cla.isOptionSet( gsdi.SDI_OPTION ) ) {
 133             base_algorithm = BASE_ALGORITHM.SDI;
 134         }
 135         if ( cla.isOptionSet( gsdi.MOST_PARSIMONIOUS_OPTION ) ) {
 136             if ( base_algorithm != BASE_ALGORITHM.GSDI ) {
 137                 ForesterUtil.fatalError( gsdi.PRG_NAME, "Can only use most parsimonious duplication mode with GSDI" );
 138             }
 139             most_parsimonous_duplication_model = true;
 140         }
 141         if ( cla.isOptionSet( gsdi.GUESS_FORMAT_OF_SPECIES_TREE ) ) {
 142             species_tree_in_phyloxml = false;
 143         }
 144         if ( cla.isOptionSet( gsdi.ALLOW_STRIPPING_OF_GENE_TREE_OPTION ) ) {
 145             if ( base_algorithm != BASE_ALGORITHM.GSDI ) {
 146                 ForesterUtil.fatalError( gsdi.PRG_NAME, "Can only allow stripping of gene tree with GSDI" );
 147             }
 148             allow_stripping_of_gene_tree = true;
 149         }
 150         Phylogeny species_tree = null;
 151         Phylogeny gene_tree = null;
 152         File gene_tree_file = null;
 153         File species_tree_file = null;
 154         File out_file = null;
 155         File log_file = null;
 156         Writer log_writer = null;
 157         try {
 158             gene_tree_file = cla.getFile( 0 );
 159             species_tree_file = cla.getFile( 1 );
 160             if ( cla.getNumberOfNames() == 3 ) {
 161                 out_file = cla.getFile( 2 );
 162             }
 163             else {
 164                 out_file = new File( ForesterUtil.removeSuffix( gene_tree_file.toString() ) + DEFAULT_OUTFILE_SUFFIX );
 165             }
 166             log_file = new File( ForesterUtil.removeSuffix( out_file.toString() ) + LOGFILE_SUFFIX );
 167         }
 168         catch ( final IllegalArgumentException e ) {
 169             ForesterUtil.fatalError( gsdi.PRG_NAME, "error in command line: " + e.getMessage() );
 170         }
 171         if ( ForesterUtil.isReadableFile( gene_tree_file ) != null ) {
 172             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isReadableFile( gene_tree_file ) );
 173         }
 174         if ( ForesterUtil.isReadableFile( species_tree_file ) != null ) {
 175             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isReadableFile( species_tree_file ) );
 176         }
 177         if ( ForesterUtil.isWritableFile( out_file ) != null ) {
 178             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isWritableFile( out_file ) );
 179         }
 180         if ( ForesterUtil.isWritableFile( log_file ) != null ) {
 181             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isWritableFile( log_file ) );
 182         }
 183         try {
 184             log_writer = ForesterUtil.createBufferedWriter( log_file );
 185         }
 186         catch ( final IOException e ) {
 187             ForesterUtil.fatalError( gsdi.PRG_NAME, "Failed to create [" + log_file + "]: " + e.getMessage() );
 188         }
 189         try {
 190             final PhylogenyFactory factory = ParserBasedPhylogenyFactory.getInstance();
 191             gene_tree = factory.create( gene_tree_file, new PhyloXmlParser() )[ 0 ];
 192         }
 193         catch ( final IOException e ) {
 194             ForesterUtil.fatalError( gsdi.PRG_NAME,
 195                                      "Failed to read gene tree from [" + gene_tree_file + "]: " + e.getMessage() );
 196         }
 197         try {
 198             final PhylogenyFactory factory = ParserBasedPhylogenyFactory.getInstance();
 199             if ( species_tree_in_phyloxml ) {
 200                 species_tree = factory.create( species_tree_file, new PhyloXmlParser() )[ 0 ];
 201             }
 202             else {
 203                 final PhylogenyParser p = ParserUtils.createParserDependingOnFileType( species_tree_file, true );
 204                 if ( REPLACE_UNDERSCORES_IN_NH_SPECIES_TREE && ( p instanceof NHXParser ) ) {
 205                     ( ( NHXParser ) p ).setReplaceUnderscores( true );
 206                 }
 207                 species_tree = factory.create( species_tree_file, p )[ 0 ];
 208                 final TaxonomyComparisonBase comp_base = GSDI.determineTaxonomyComparisonBase( gene_tree );
 209                 switch ( comp_base ) {
 210                     case SCIENTIFIC_NAME:
 211                         PhylogenyMethods
 212                                 .transferNodeNameToField( species_tree,
 213                                                           PhylogenyMethods.PhylogenyNodeField.TAXONOMY_ID_UNIPROT_1 );
 214                         break;
 215                     case CODE:
 216                         PhylogenyMethods.transferNodeNameToField( species_tree,
 217                                                                   PhylogenyMethods.PhylogenyNodeField.TAXONOMY_CODE );
 218                         break;
 219                     case ID:
 220                         PhylogenyMethods.transferNodeNameToField( species_tree,
 221                                                                   PhylogenyMethods.PhylogenyNodeField.TAXONOMY_ID );
 222                         break;
 223                     default:
 224                         ForesterUtil.fatalError( gsdi.PRG_NAME, "unable to determine comparison base" );
 225                 }
 226             }
 227         }
 228         catch ( final IOException e ) {
 229             ForesterUtil.fatalError( gsdi.PRG_NAME,
 230                                      "Failed to read species tree from [" + gene_tree_file + "]: " + e.getMessage() );
 231         }
 232         gene_tree.setRooted( true );
 233         species_tree.setRooted( true );
 234         if ( !gene_tree.isCompletelyBinary() ) {
 235             log_writer.write( "User Error: gene tree is not completely binary" );
 236             log_writer.write( ForesterUtil.LINE_SEPARATOR );
 237             log_writer.close();
 238             ForesterUtil.fatalError( gsdi.PRG_NAME, "gene tree is not completely binary" );
 239         }
 240         if ( base_algorithm != BASE_ALGORITHM.GSDI ) {
 241             if ( !species_tree.isCompletelyBinary() ) {
 242                 log_writer.write( "User Error: species tree is not completely binary, use GSDI instead" );
 243                 log_writer.write( ForesterUtil.LINE_SEPARATOR );
 244                 log_writer.close();
 245                 ForesterUtil.fatalError( gsdi.PRG_NAME, "species tree is not completely binary, use GSDI instead" );
 246             }
 247         }
 248         // For timing.
 249         // gene_tree = Helper.createBalancedTree( 10 );
 250         // species_tree = Helper.createBalancedTree( 13 );
 251         // species_tree = Helper.createUnbalancedTree( 1024 );
 252         // gene_tree = Helper.createUnbalancedTree( 8192 );
 253         // species_tree = gene_tree.copyTree();
 254         // gene_tree = species_tree.copyTree();
 255         // Helper.numberSpeciesInOrder( species_tree );
 256         // Helper.numberSpeciesInOrder( gene_tree );
 257         // Helper.randomizeSpecies( 1, 8192, gene_tree );
 258         // Helper.intervalNumberSpecies( gene_tree, 4096 );
 259         // Helper.numberSpeciesInDescOrder( gene_tree );
 260         log_writer.write( PRG_NAME + " - " + PRG_DESC );
 261         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 262         log_writer.write( "  version         : " + PRG_VERSION );
 263         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 264         log_writer.write( "  date            : " + PRG_DATE );
 265         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 266         log_writer.write( "  forester version: " + ForesterConstants.FORESTER_VERSION );
 267         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 268         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 269         log_writer.write( "Start time: " + new SimpleDateFormat( "yyyyMMdd HH:mm:ss" ).format( new Date() ) );
 270         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 271         log_writer.write( "Gene tree file: " + gene_tree_file.getCanonicalPath() );
 272         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 273         log_writer.write( "Gene tree name: "
 274                 + ( ForesterUtil.isEmpty( gene_tree.getName() ) ? "" : gene_tree.getName() ) );
 275         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 276         log_writer.write( "Species tree file: " + species_tree_file.getCanonicalPath() );
 277         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 278         log_writer.write( "Species tree name: "
 279                 + ( ForesterUtil.isEmpty( species_tree.getName() ) ? "" : gene_tree.getName() ) );
 280         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 281         System.out.println();
 282         System.out.println( "Strip species tree: " + strip_species_tree );
 283         log_writer.write( "Strip species tree: " + strip_species_tree );
 284         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 285         SDI sdi = null;
 286         final long start_time = new Date().getTime();
 287         try {
 288             if ( base_algorithm == BASE_ALGORITHM.GSDI ) {
 289                 System.out.println();
 290                 System.out.println( "Use most parsimonous duplication model: " + most_parsimonous_duplication_model );
 291                 System.out.println( "Allow stripping of gene tree nodes    : " + allow_stripping_of_gene_tree );
 292                 log_writer.write( "Use most parsimonous duplication model: " + most_parsimonous_duplication_model );
 293                 log_writer.write( ForesterUtil.LINE_SEPARATOR );
 294                 log_writer.write( "Allow stripping of gene tree nodes    : " + allow_stripping_of_gene_tree );
 295                 log_writer.write( ForesterUtil.LINE_SEPARATOR );
 296                 log_writer.flush();
 297                 sdi = new GSDI( gene_tree,
 298                                 species_tree,
 299                                 most_parsimonous_duplication_model,
 300                                 allow_stripping_of_gene_tree,
 301                                 strip_species_tree );
 302             }
 303             else {
 304                 System.out.println();
 305                 System.out.println( "Using SDIse algorithm" );
 306                 log_writer.write( "Using SDIse algorithm" );
 307                 log_writer.write( ForesterUtil.LINE_SEPARATOR );
 308                 log_writer.flush();
 309                 sdi = new SDIse( gene_tree, species_tree );
 310             }
 311         }
 312         catch ( final SdiException e ) {
 313             log_writer.write( "User Error: " + e.getLocalizedMessage() );
 314             log_writer.write( ForesterUtil.LINE_SEPARATOR );
 315             log_writer.close();
 316             ForesterUtil.fatalError( PRG_NAME, e.getLocalizedMessage() );
 317         }
 318         catch ( final IOException e ) {
 319             log_writer.write( "Error: " + e );
 320             log_writer.write( ForesterUtil.LINE_SEPARATOR );
 321             log_writer.close();
 322             ForesterUtil.fatalError( PRG_NAME, e.toString() );
 323         }
 324         catch ( final Exception e ) {
 325             log_writer.write( "Error: " + e );
 326             log_writer.write( ForesterUtil.LINE_SEPARATOR );
 327             log_writer.close();
 328             e.printStackTrace();
 329             System.exit( -1 );
 330         }
 331         System.out.println();
 332         System.out.println( "Running time (excluding I/O): " + ( new Date().getTime() - start_time ) + "ms" );
 333         log_writer.write( "Running time (excluding I/O): " + ( new Date().getTime() - start_time ) + "ms" );
 334         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 335         try {
 336             final PhylogenyWriter writer = new PhylogenyWriter();
 337             writer.toPhyloXML( out_file, gene_tree, 0 );
 338         }
 339         catch ( final IOException e ) {
 340             ForesterUtil.fatalError( PRG_NAME, "Failed to write to [" + out_file + "]: " + e.getMessage() );
 341         }
 342         System.out.println();
 343         System.out.println( "Successfully wrote resulting gene tree to: " + out_file );
 344         System.out.println();
 345         log_writer.write( "Wrote resulting gene tree to: " + out_file );
 346         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 347         if ( base_algorithm == BASE_ALGORITHM.SDI ) {
 348             sdi.computeMappingCostL();
 349             System.out.println( "Mapping cost                    : " + sdi.computeMappingCostL() );
 350             log_writer.write( "Mapping cost                    : " + sdi.computeMappingCostL() );
 351             log_writer.write( ForesterUtil.LINE_SEPARATOR );
 352         }
 353         System.out.println( "Number of duplications          : " + sdi.getDuplicationsSum() );
 354         log_writer.write( "Number of duplications          : " + sdi.getDuplicationsSum() );
 355         log_writer.write( ForesterUtil.LINE_SEPARATOR );
 356         if ( ( base_algorithm == BASE_ALGORITHM.GSDI ) ) {
 357             final GSDI gsdi = ( GSDI ) sdi;
 358             if ( !most_parsimonous_duplication_model ) {
 359                 final int duplications = gsdi.getSpeciationOrDuplicationEventsSum();
 360                 System.out.println( "Number of potential duplications: " + duplications );
 361                 log_writer.write( "Number of potential duplications: " + duplications );
 362                 log_writer.write( ForesterUtil.LINE_SEPARATOR );
 363             }
 364             final int spec = gsdi.getSpeciationsSum();
 365             System.out.println( "Number of speciations            : " + spec );
 366             log_writer.write( "Number of speciations            : " + spec );
 367             log_writer.write( ForesterUtil.LINE_SEPARATOR );
 368             for( PhylogenyNode n : gsdi.getMappedExternalSpeciesTreeNodes() ) {
 369                 System.out.println( n.toString() );
 370             }
 371         }
 372         System.out.println();
 373         log_writer.close();
 374         //  some stat on gene tree:
 375         //      filename, name
 376         //      number of external nodes, strppided nodes
 377         //  some stats on sepcies tree, external nodes,
 378         //  filename, name
 379         //  internal nodes
 380         //  how many of which are polytomies
 381         //wrote log file to
 382         //  if ( allow_stripping_of_gene_tree ) {
 383         //      stripped x nodes, y external nodes remain
 384         //  }
 385     }
 386
 387     private static void print_help() {
 388         System.out.println( "Usage: " + gsdi.PRG_NAME
 389                 + " [-options] <gene tree in phyloXML format> <species tree in phyloXML format> [outfile]" );
 390         System.out.println();
 391         System.out.println( "Options:" );
 392         //    System.out.println( " -" + gsdi.ALLOW_STRIPPING_OF_GENE_TREE_OPTION + ": to allow stripping of gene tree nodes without a matching species in the species tree (writes list of stripped nodes to " + );
 393         System.out.println( " -" + gsdi.STRIP_OPTION
 394                 + ": to strip the species tree of unneeded species prior to duplication inference" );
 395         System.out.println( " -" + gsdi.SDI_OPTION + ": to use SDI algorithm instead of GSDI algorithm" );//TODO gsdi.ALLOW_STRIPPING_OF_GENE_TREE_OPTION not allowed
 396         System.out.println( " -" + gsdi.MOST_PARSIMONIOUS_OPTION
 397                 + ": use most parimonious duplication model for GSDI: " );
 398         System.out.println( "     assign nodes as speciations which would otherwise be assiged" );
 399         System.out.println( "     as unknown because of polytomies in the species tree" );
 400         System.out.println( " -" + gsdi.GUESS_FORMAT_OF_SPECIES_TREE
 401                 + ": to allow species tree in other formats than phyloXML (i.e. Newick, NHX, Nexus)" );
 402         System.out.println();
 403         System.out.println( "Species tree:" );
 404         System.out
 405                 .println( " In phyloXML format (unless option -"
 406                         + gsdi.GUESS_FORMAT_OF_SPECIES_TREE
 407                         + " is used, in which case the species matching between gene tree and species tree must be via scientific name), with taxonomy data in appropriate fields" );
 408         System.out.println();
 409         System.out.println( "Gene tree:" );
 410         System.out.println( " In phyloXM format, with taxonomy and sequence data in appropriate fields" );
 411         System.out.println();
 412         System.out.println( "Example:" );
 413         //  System.out.println( "gsdi
 414         //    System.out.println();
 415         System.out.println( "Note -- GSDI algorithm is under development" );
 416         System.out.println();
 417     }
 418 }