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.text.SimpleDateFormat;
  31 import java.util.ArrayList;
  32 import java.util.Date;
  33 import java.util.List;
  34
  35 import org.forester.io.parsers.PhylogenyParser;
  36 import org.forester.io.parsers.nhx.NHXParser;
  37 import org.forester.io.parsers.phyloxml.PhyloXmlParser;
  38 import org.forester.io.parsers.util.ParserUtils;
  39 import org.forester.io.writers.PhylogenyWriter;
  40 import org.forester.phylogeny.Phylogeny;
  41 import org.forester.phylogeny.PhylogenyMethods;
  42 import org.forester.phylogeny.PhylogenyNode;
  43 import org.forester.phylogeny.factories.ParserBasedPhylogenyFactory;
  44 import org.forester.phylogeny.factories.PhylogenyFactory;
  45 import org.forester.sdi.GSDI;
  46 import org.forester.sdi.SDI;
  47 import org.forester.sdi.SDI.TaxonomyComparisonBase;
  48 import org.forester.sdi.SDIse;
  49 import org.forester.sdi.SdiException;
  50 import org.forester.util.CommandLineArguments;
  51 import org.forester.util.EasyWriter;
  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 ALLOW_STRIPPING_OF_GENE_TREE_OPTION        = "g";
  62     final static private String SDI_OPTION                                 = "b";
  63     final static private String MOST_PARSIMONIOUS_OPTION                   = "m";
  64     final static private String GUESS_FORMAT_OF_SPECIES_TREE               = "q";
  65     final static private String HELP_OPTION_1                              = "help";
  66     final static private String HELP_OPTION_2                              = "h";
  67     final static private String DEFAULT_OUTFILE_SUFFIX                     = "_gsdi_out.xml";
  68     final static private String SUFFIX_FOR_LIST_OF_STIPPED_GENE_TREE_NODES = "_stripped_gene_tree_nodes.txt";
  69     final static private String SUFFIX_FOR_SPECIES_TREE_USED               = "_species_tree_used.xml";
  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.SDI_OPTION );
 108             allowed_options.add( gsdi.GUESS_FORMAT_OF_SPECIES_TREE );
 109             allowed_options.add( gsdi.MOST_PARSIMONIOUS_OPTION );
 110             allowed_options.add( gsdi.ALLOW_STRIPPING_OF_GENE_TREE_OPTION );
 111             final String dissallowed_options = cla.validateAllowedOptionsAsString( allowed_options );
 112             if ( dissallowed_options.length() > 0 ) {
 113                 ForesterUtil.fatalError( gsdi.PRG_NAME, "unknown option(s): " + dissallowed_options );
 114             }
 115             execute( cla );
 116         }
 117         catch ( final IOException e ) {
 118             ForesterUtil.fatalError( gsdi.PRG_NAME, e.getMessage() );
 119         }
 120     }
 121
 122     private static void execute( final CommandLineArguments cla ) throws IOException {
 123         BASE_ALGORITHM base_algorithm = BASE_ALGORITHM.GSDI;
 124         boolean most_parsimonous_duplication_model = false;
 125         boolean species_tree_in_phyloxml = true;
 126         boolean allow_stripping_of_gene_tree = false;
 127         if ( cla.isOptionSet( gsdi.SDI_OPTION ) ) {
 128             base_algorithm = BASE_ALGORITHM.SDI;
 129         }
 130         if ( cla.isOptionSet( gsdi.MOST_PARSIMONIOUS_OPTION ) ) {
 131             if ( base_algorithm != BASE_ALGORITHM.GSDI ) {
 132                 ForesterUtil.fatalError( gsdi.PRG_NAME, "Can only use most parsimonious duplication mode with GSDI" );
 133             }
 134             most_parsimonous_duplication_model = true;
 135         }
 136         if ( cla.isOptionSet( gsdi.GUESS_FORMAT_OF_SPECIES_TREE ) ) {
 137             species_tree_in_phyloxml = false;
 138         }
 139         if ( cla.isOptionSet( gsdi.ALLOW_STRIPPING_OF_GENE_TREE_OPTION ) ) {
 140             if ( base_algorithm != BASE_ALGORITHM.GSDI ) {
 141                 ForesterUtil.fatalError( gsdi.PRG_NAME, "Can only allow stripping of gene tree with GSDI" );
 142             }
 143             allow_stripping_of_gene_tree = true;
 144         }
 145         Phylogeny species_tree = null;
 146         Phylogeny gene_tree = null;
 147         File gene_tree_file = null;
 148         File species_tree_file = null;
 149         File out_file = null;
 150         File log_file = null;
 151         EasyWriter log_writer = null;
 152         try {
 153             gene_tree_file = cla.getFile( 0 );
 154             species_tree_file = cla.getFile( 1 );
 155             if ( cla.getNumberOfNames() == 3 ) {
 156                 out_file = cla.getFile( 2 );
 157             }
 158             else {
 159                 out_file = new File( ForesterUtil.removeSuffix( gene_tree_file.toString() ) + DEFAULT_OUTFILE_SUFFIX );
 160             }
 161             log_file = new File( ForesterUtil.removeSuffix( out_file.toString() ) + LOGFILE_SUFFIX );
 162         }
 163         catch ( final IllegalArgumentException e ) {
 164             ForesterUtil.fatalError( gsdi.PRG_NAME, "error in command line: " + e.getMessage() );
 165         }
 166         if ( ForesterUtil.isReadableFile( gene_tree_file ) != null ) {
 167             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isReadableFile( gene_tree_file ) );
 168         }
 169         if ( ForesterUtil.isReadableFile( species_tree_file ) != null ) {
 170             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isReadableFile( species_tree_file ) );
 171         }
 172         if ( ForesterUtil.isWritableFile( out_file ) != null ) {
 173             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isWritableFile( out_file ) );
 174         }
 175         if ( ForesterUtil.isWritableFile( log_file ) != null ) {
 176             ForesterUtil.fatalError( gsdi.PRG_NAME, ForesterUtil.isWritableFile( log_file ) );
 177         }
 178         try {
 179             log_writer = ForesterUtil.createEasyWriter( log_file );
 180         }
 181         catch ( final IOException e ) {
 182             ForesterUtil.fatalError( gsdi.PRG_NAME, "Failed to create [" + log_file + "]: " + e.getMessage() );
 183         }
 184         try {
 185             final PhylogenyFactory factory = ParserBasedPhylogenyFactory.getInstance();
 186             gene_tree = factory.create( gene_tree_file, new PhyloXmlParser() )[ 0 ];
 187         }
 188         catch ( final IOException e ) {
 189             ForesterUtil.fatalError( gsdi.PRG_NAME,
 190                                      "Failed to read gene tree from [" + gene_tree_file + "]: " + e.getMessage() );
 191         }
 192         try {
 193             final PhylogenyFactory factory = ParserBasedPhylogenyFactory.getInstance();
 194             if ( species_tree_in_phyloxml ) {
 195                 species_tree = factory.create( species_tree_file, new PhyloXmlParser() )[ 0 ];
 196             }
 197             else {
 198                 final PhylogenyParser p = ParserUtils.createParserDependingOnFileType( species_tree_file, true );
 199                 if ( REPLACE_UNDERSCORES_IN_NH_SPECIES_TREE && ( p instanceof NHXParser ) ) {
 200                     ( ( NHXParser ) p ).setReplaceUnderscores( true );
 201                 }
 202                 species_tree = factory.create( species_tree_file, p )[ 0 ];
 203                 final TaxonomyComparisonBase comp_base = GSDI.determineTaxonomyComparisonBase( gene_tree );
 204                 switch ( comp_base ) {
 205                     case SCIENTIFIC_NAME:
 206                         PhylogenyMethods
 207                                 .transferNodeNameToField( species_tree,
 208                                                           PhylogenyMethods.PhylogenyNodeField.TAXONOMY_ID_UNIPROT_1,
 209                                                           true );
 210                         break;
 211                     case CODE:
 212                         PhylogenyMethods.transferNodeNameToField( species_tree,
 213                                                                   PhylogenyMethods.PhylogenyNodeField.TAXONOMY_CODE,
 214                                                                   true );
 215                         break;
 216                     case ID:
 217                         PhylogenyMethods.transferNodeNameToField( species_tree,
 218                                                                   PhylogenyMethods.PhylogenyNodeField.TAXONOMY_ID,
 219                                                                   true );
 220                         break;
 221                     default:
 222                         ForesterUtil.fatalError( gsdi.PRG_NAME, "unable to determine comparison base" );
 223                 }
 224             }
 225         }
 226         catch ( final IOException e ) {
 227             ForesterUtil.fatalError( gsdi.PRG_NAME,
 228                                      "Failed to read species tree from [" + gene_tree_file + "]: " + e.getMessage() );
 229         }
 230         gene_tree.setRooted( true );
 231         species_tree.setRooted( true );
 232         if ( !gene_tree.isCompletelyBinary() ) {
 233             log_writer.println( "User Error: gene tree is not completely binary" );
 234             log_writer.close();
 235             ForesterUtil.fatalError( gsdi.PRG_NAME, "gene tree is not completely binary" );
 236         }
 237         if ( base_algorithm != BASE_ALGORITHM.GSDI ) {
 238             if ( !species_tree.isCompletelyBinary() ) {
 239                 log_writer.println( "User Error: species tree is not completely binary, use GSDI instead" );
 240                 log_writer.close();
 241                 ForesterUtil.fatalError( gsdi.PRG_NAME, "species tree is not completely binary, use GSDI instead" );
 242             }
 243         }
 244         // For timing.
 245         // gene_tree = Helper.createBalancedTree( 10 );
 246         // species_tree = Helper.createBalancedTree( 13 );
 247         // species_tree = Helper.createUnbalancedTree( 1024 );
 248         // gene_tree = Helper.createUnbalancedTree( 8192 );
 249         // species_tree = gene_tree.copyTree();
 250         // gene_tree = species_tree.copyTree();
 251         // Helper.numberSpeciesInOrder( species_tree );
 252         // Helper.numberSpeciesInOrder( gene_tree );
 253         // Helper.randomizeSpecies( 1, 8192, gene_tree );
 254         // Helper.intervalNumberSpecies( gene_tree, 4096 );
 255         // Helper.numberSpeciesInDescOrder( gene_tree );
 256         log_writer.println( PRG_NAME + " - " + PRG_DESC );
 257         log_writer.println( "  version         : " + PRG_VERSION );
 258         log_writer.println( "  date            : " + PRG_DATE );
 259         log_writer.println( "  forester version: " + ForesterConstants.FORESTER_VERSION );
 260         log_writer.println( "Start time: " + new SimpleDateFormat( "yyyyMMdd HH:mm:ss" ).format( new Date() ) );
 261         log_writer.println( "Gene tree file: " + gene_tree_file.getCanonicalPath() );
 262         log_writer.println( "Gene tree name: "
 263                 + ( ForesterUtil.isEmpty( gene_tree.getName() ) ? "" : gene_tree.getName() ) );
 264         log_writer.println( "Species tree file: " + species_tree_file.getCanonicalPath() );
 265         log_writer.println( "Species tree name: "
 266                 + ( ForesterUtil.isEmpty( species_tree.getName() ) ? "" : gene_tree.getName() ) );
 267         System.out.println();
 268         SDI sdi = null;
 269         final long start_time = new Date().getTime();
 270         try {
 271             if ( base_algorithm == BASE_ALGORITHM.GSDI ) {
 272                 System.out.println();
 273                 System.out.println( "Use most parsimonous duplication model: " + most_parsimonous_duplication_model );
 274                 System.out.println( "Allow stripping of gene tree nodes    : " + allow_stripping_of_gene_tree );
 275                 log_writer.println( "Use most parsimonous duplication model: " + most_parsimonous_duplication_model );
 276                 log_writer.write( "Allow stripping of gene tree nodes    : " + allow_stripping_of_gene_tree );
 277                 log_writer.flush();
 278                 sdi = new GSDI( gene_tree,
 279                                 species_tree,
 280                                 most_parsimonous_duplication_model,
 281                                 allow_stripping_of_gene_tree,
 282                                 true );
 283             }
 284             else {
 285                 System.out.println();
 286                 System.out.println( "Using SDIse algorithm" );
 287                 log_writer.println( "Using SDIse algorithm" );
 288                 log_writer.flush();
 289                 sdi = new SDIse( gene_tree, species_tree );
 290             }
 291         }
 292         catch ( final SdiException e ) {
 293             log_writer.println( "User Error: " + e.getLocalizedMessage() );
 294             log_writer.close();
 295             ForesterUtil.fatalError( PRG_NAME, e.getLocalizedMessage() );
 296         }
 297         catch ( final IOException e ) {
 298             log_writer.println( "Error: " + e );
 299             log_writer.close();
 300             ForesterUtil.fatalError( PRG_NAME, e.toString() );
 301         }
 302         catch ( final Exception e ) {
 303             log_writer.println( "Error: " + e );
 304             log_writer.close();
 305             e.printStackTrace();
 306             System.exit( -1 );
 307         }
 308         System.out.println();
 309         System.out.println( "Running time (excluding I/O): " + ( new Date().getTime() - start_time ) + "ms" );
 310         log_writer.println( "Running time (excluding I/O): " + ( new Date().getTime() - start_time ) + "ms" );
 311         try {
 312             final PhylogenyWriter writer = new PhylogenyWriter();
 313             writer.toPhyloXML( out_file, gene_tree, 0 );
 314         }
 315         catch ( final IOException e ) {
 316             ForesterUtil.fatalError( PRG_NAME, "Failed to write to [" + out_file + "]: " + e.getMessage() );
 317         }
 318         System.out.println();
 319         System.out.println( "Wrote resulting gene tree to: " + out_file );
 320         System.out.println();
 321         log_writer.println( "Wrote resulting gene tree to: " + out_file );
 322         if ( base_algorithm == BASE_ALGORITHM.SDI ) {
 323             sdi.computeMappingCostL();
 324             System.out.println( "Mapping cost                    : " + sdi.computeMappingCostL() );
 325             log_writer.println( "Mapping cost                    : " + sdi.computeMappingCostL() );
 326         }
 327         System.out.println( "Number of duplications          : " + sdi.getDuplicationsSum() );
 328         log_writer.println( "Number of duplications          : " + sdi.getDuplicationsSum() );
 329         if ( ( base_algorithm == BASE_ALGORITHM.GSDI ) ) {
 330             final GSDI gsdi = ( GSDI ) sdi;
 331             final File species_tree_used_file = new File( out_file + SUFFIX_FOR_SPECIES_TREE_USED );
 332             try {
 333                 final PhylogenyWriter writer = new PhylogenyWriter();
 334                 writer.toPhyloXML( species_tree_used_file, gsdi.getSpeciesTree(), 0 );
 335             }
 336             catch ( final IOException e ) {
 337                 ForesterUtil.fatalError( PRG_NAME,
 338                                          "Failed to write to [" + species_tree_used_file + "]: " + e.getMessage() );
 339             }
 340             System.out.println();
 341             System.out.println( "Wrote used species tree to: " + species_tree_used_file );
 342             System.out.println();
 343             log_writer.println( "Wrote used species tree to: " + species_tree_used_file );
 344             if ( !most_parsimonous_duplication_model ) {
 345                 final int duplications = gsdi.getSpeciationOrDuplicationEventsSum();
 346                 System.out.println( "Number of potential duplications: " + duplications );
 347                 log_writer.println( "Number of potential duplications: " + duplications );
 348             }
 349             final int spec = gsdi.getSpeciationsSum();
 350             System.out.println( "Number of speciations            : " + spec );
 351             log_writer.println( "Number of speciations            : " + spec );
 352             for( final PhylogenyNode n : gsdi.getMappedExternalSpeciesTreeNodes() ) {
 353                 System.out.println( n.toString() );
 354             }
 355         }
 356         System.out.println();
 357         log_writer.close();
 358         //  some stat on gene tree:
 359         //      filename, name
 360         //      number of external nodes, strppided nodes
 361         //  some stats on sepcies tree, external nodes,
 362         //  filename, name
 363         //  internal nodes
 364         //  how many of which are polytomies
 365         //wrote log file to
 366         //  if ( allow_stripping_of_gene_tree ) {
 367         //      stripped x nodes, y external nodes remain
 368         //  }
 369     }
 370
 371     private static void print_help() {
 372         System.out.println( "Usage: " + gsdi.PRG_NAME
 373                 + " [-options] <gene tree in phyloXML format> <species tree in phyloXML format> [outfile]" );
 374         System.out.println();
 375         System.out.println( "Options:" );
 376         //    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 " + );
 377         System.out.println( " -" + gsdi.SDI_OPTION + ": to use SDI algorithm instead of GSDI algorithm" );//TODO gsdi.ALLOW_STRIPPING_OF_GENE_TREE_OPTION not allowed
 378         System.out.println( " -" + gsdi.MOST_PARSIMONIOUS_OPTION
 379                 + ": use most parimonious duplication model for GSDI: " );
 380         System.out.println( "     assign nodes as speciations which would otherwise be assiged" );
 381         System.out.println( "     as unknown because of polytomies in the species tree" );
 382         System.out.println( " -" + gsdi.GUESS_FORMAT_OF_SPECIES_TREE
 383                 + ": to allow species tree in other formats than phyloXML (i.e. Newick, NHX, Nexus)" );
 384         System.out.println();
 385         System.out.println( "Species tree:" );
 386         System.out
 387                 .println( " In phyloXML format (unless option -"
 388                         + gsdi.GUESS_FORMAT_OF_SPECIES_TREE
 389                         + " 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" );
 390         System.out.println();
 391         System.out.println( "Gene tree:" );
 392         System.out.println( " In phyloXM format, with taxonomy and sequence data in appropriate fields" );
 393         System.out.println();
 394         System.out.println( "Example:" );
 395         //  System.out.println( "gsdi
 396         //    System.out.println();
 397         System.out.println( "Note -- GSDI algorithm is under development" );
 398         System.out.println();
 399     }
 400 }