forester/java/src/org/forester/surfacing/PairwiseGenomeComparator.java

   1 // $Id:
   2 //
   3 // FORESTER -- software libraries and applications
   4 // for evolutionary biology research and applications.
   5 //
   6 // Copyright (C) 2008-2009 Christian M. Zmasek
   7 // Copyright (C) 2008-2009 Burnham Institute for Medical Research
   8 // All rights reserved
   9 //
  10 // This library is free software; you can redistribute it and/or
  11 // modify it under the terms of the GNU Lesser General Public
  12 // License as published by the Free Software Foundation; either
  13 // version 2.1 of the License, or (at your option) any later version.
  14 //
  15 // This library is distributed in the hope that it will be useful,
  16 // but WITHOUT ANY WARRANTY; without even the implied warranty of
  17 // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  18 // Lesser General Public License for more details.
  19 //
  20 // You should have received a copy of the GNU Lesser General Public
  21 // License along with this library; if not, write to the Free Software
  22 // Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA
  23 //
  24 // Contact: phylosoft @ gmail . com
  25 // WWW: https://sites.google.com/site/cmzmasek/home/software/forester
  26
  27 package org.forester.surfacing;
  28
  29 import java.io.BufferedWriter;
  30 import java.io.File;
  31 import java.io.FileWriter;
  32 import java.io.IOException;
  33 import java.io.Writer;
  34 import java.util.ArrayList;
  35 import java.util.List;
  36 import java.util.Map;
  37 import java.util.Random;
  38 import java.util.SortedSet;
  39 import java.util.TreeSet;
  40
  41 import org.forester.evoinference.matrix.distance.BasicSymmetricalDistanceMatrix;
  42 import org.forester.evoinference.matrix.distance.DistanceMatrix;
  43 import org.forester.go.GoId;
  44 import org.forester.go.GoNameSpace;
  45 import org.forester.go.GoTerm;
  46 import org.forester.protein.DomainId;
  47 import org.forester.species.Species;
  48 import org.forester.surfacing.DomainSimilarityCalculator.Detailedness;
  49 import org.forester.util.DescriptiveStatistics;
  50 import org.forester.util.ForesterUtil;
  51
  52 public class PairwiseGenomeComparator {
  53
  54     private List<DistanceMatrix> _domain_distance_scores_means;
  55     private List<DistanceMatrix> _shared_domains_based_distances;
  56     private List<DistanceMatrix> _shared_binary_combinations_based_distances;
  57
  58     //private List<HistogramData>  _histogram_datas;
  59     public PairwiseGenomeComparator() {
  60         init();
  61     }
  62
  63     public List<DistanceMatrix> getDomainDistanceScoresMeans() {
  64         return _domain_distance_scores_means;
  65     }
  66
  67     //public List<HistogramData> getHistogramDatas() {
  68     //    return _histogram_datas;
  69     //}
  70     public List<DistanceMatrix> getSharedBinaryCombinationsBasedDistances() {
  71         return _shared_binary_combinations_based_distances;
  72     }
  73
  74     public List<DistanceMatrix> getSharedDomainsBasedDistances() {
  75         return _shared_domains_based_distances;
  76     }
  77
  78     private void init() {
  79         //_histogram_datas = new ArrayList<HistogramData>();
  80         _domain_distance_scores_means = new ArrayList<DistanceMatrix>();
  81         _shared_domains_based_distances = new ArrayList<DistanceMatrix>();
  82         _shared_binary_combinations_based_distances = new ArrayList<DistanceMatrix>();
  83     }
  84
  85     public void performPairwiseComparisons( final StringBuilder html_desc,
  86                                             final boolean sort_by_species_count_first,
  87                                             final Detailedness detailedness,
  88                                             final boolean ignore_domains_without_combs_in_all_spec,
  89                                             final boolean ignore_domains_specific_to_one_species,
  90                                             final DomainSimilarity.DomainSimilaritySortField domain_similarity_sort_field,
  91                                             final PrintableDomainSimilarity.PRINT_OPTION domain_similarity_print_option,
  92                                             final DomainSimilarity.DomainSimilarityScoring scoring,
  93                                             final Map<DomainId, List<GoId>> domain_id_to_go_ids_map,
  94                                             final Map<GoId, GoTerm> go_id_to_term_map,
  95                                             final GoNameSpace go_namespace_limit,
  96                                             final Species[] species,
  97                                             final int number_of_genomes,
  98                                             final List<GenomeWideCombinableDomains> list_of_genome_wide_combinable_domains,
  99                                             final PairwiseDomainSimilarityCalculator pw_calc,
 100                                             final String automated_pairwise_comparison_suffix,
 101                                             final boolean verbose,
 102                                             final String automated_pairwise_comparison_prefix,
 103                                             final String command_line_prg_name,
 104                                             final File out_dir,
 105                                             final boolean write_pairwise_comparisons,
 106                                             final Map<String, Integer> tax_code_to_id_map ) {
 107         init();
 108         final BasicSymmetricalDistanceMatrix domain_distance_scores_means = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 109         final BasicSymmetricalDistanceMatrix shared_domains_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 110         final BasicSymmetricalDistanceMatrix shared_binary_combinations_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 111         if ( verbose ) {
 112             System.out.println();
 113             System.out.println( "Pairwise genome distances:" );
 114             System.out.print( "[species-i - species-j:" );
 115             System.out.print( " mean-score-based" );
 116             System.out.print( " (sd)" );
 117             System.out.print( " [N]" );
 118             System.out.print( " | shared-domains-based" );
 119             System.out.println( " | shared-binary-combinations-based]" );
 120             System.out.println();
 121         }
 122         for( int i = 0; i < number_of_genomes; ++i ) {
 123             final String species_i = species[ i ].getSpeciesId();
 124             domain_distance_scores_means.setIdentifier( i, species_i );
 125             shared_domains_based_distances.setIdentifier( i, species_i );
 126             shared_binary_combinations_based_distances.setIdentifier( i, species_i );
 127             if ( verbose ) {
 128                 System.out.println( ( i + 1 ) + "/" + number_of_genomes );
 129             }
 130             for( int j = 0; j < i; ++j ) {
 131                 if ( ( list_of_genome_wide_combinable_domains.get( i ).getSize() < 1 )
 132                         || ( list_of_genome_wide_combinable_domains.get( j ).getSize() < 1 ) ) {
 133                     domain_distance_scores_means
 134                             .setValue( i, j, DomainArchitectureBasedGenomeSimilarityCalculator.MAX_SIMILARITY_SCORE );
 135                     shared_domains_based_distances
 136                             .setValue( i, j, DomainArchitectureBasedGenomeSimilarityCalculator.MAX_SIMILARITY_SCORE );
 137                     shared_binary_combinations_based_distances
 138                             .setValue( i, j, DomainArchitectureBasedGenomeSimilarityCalculator.MAX_SIMILARITY_SCORE );
 139                     continue;
 140                 }
 141                 final List<GenomeWideCombinableDomains> genome_pair = new ArrayList<GenomeWideCombinableDomains>( 2 );
 142                 genome_pair.add( list_of_genome_wide_combinable_domains.get( i ) );
 143                 genome_pair.add( list_of_genome_wide_combinable_domains.get( j ) );
 144                 DomainSimilarityCalculator.GoAnnotationOutput go_annotation_output = DomainSimilarityCalculator.GoAnnotationOutput.NONE;
 145                 if ( domain_id_to_go_ids_map != null ) {
 146                     go_annotation_output = DomainSimilarityCalculator.GoAnnotationOutput.ALL;
 147                 }
 148                 final DomainSimilarityCalculator calc = new BasicDomainSimilarityCalculator( domain_similarity_sort_field,
 149                                                                                              sort_by_species_count_first,
 150                                                                                              true );
 151                 final SortedSet<DomainSimilarity> similarities = calc
 152                         .calculateSimilarities( pw_calc,
 153                                                 genome_pair,
 154                                                 ignore_domains_without_combs_in_all_spec,
 155                                                 ignore_domains_specific_to_one_species );
 156                 SurfacingUtil.decoratePrintableDomainSimilarities( similarities,
 157                                                                    detailedness,
 158                                                                    go_annotation_output,
 159                                                                    go_id_to_term_map,
 160                                                                    go_namespace_limit );
 161                 final DescriptiveStatistics stats = SurfacingUtil
 162                         .calculateDescriptiveStatisticsForMeanValues( similarities );
 163                 final String species_j = species[ j ].getSpeciesId();
 164                 final DomainArchitectureBasedGenomeSimilarityCalculator genome_similarity_calculator = new DomainArchitectureBasedGenomeSimilarityCalculator( list_of_genome_wide_combinable_domains
 165                                                                                                                                                                       .get( i ),
 166                                                                                                                                                               list_of_genome_wide_combinable_domains
 167                                                                                                                                                                       .get( j ) );
 168                 genome_similarity_calculator.setAllowDomainsToBeIgnored( false );
 169                 double dissimilarity_score_mean;
 170                 if ( stats.getN() < 1 ) {
 171                     // No domains in common
 172                     dissimilarity_score_mean = 1.0;
 173                 }
 174                 else {
 175                     dissimilarity_score_mean = 1.0 - stats.arithmeticMean();
 176                 }
 177                 final double shared_domains_based_genome_distance = 1.0 - genome_similarity_calculator
 178                         .calculateSharedDomainsBasedGenomeSimilarityScore();
 179                 final double shared_binary_combinations_based_genome_distance = 1.0 - genome_similarity_calculator
 180                         .calculateSharedBinaryDomainCombinationBasedGenomeSimilarityScore();
 181                 domain_distance_scores_means.setValue( i, j, dissimilarity_score_mean );
 182                 shared_domains_based_distances.setValue( i, j, shared_domains_based_genome_distance );
 183                 shared_binary_combinations_based_distances.setValue( i,
 184                                                                      j,
 185                                                                      shared_binary_combinations_based_genome_distance );
 186                 if ( verbose ) {
 187                     System.out.print( species_i + "-" );
 188                     System.out.print( species_j + ": " );
 189                     System.out.print( ForesterUtil.round( dissimilarity_score_mean, 2 ) );
 190                     if ( stats.getN() > 1 ) {
 191                         System.out.print( " (" + ForesterUtil.round( stats.sampleStandardDeviation(), 2 ) + ")" );
 192                     }
 193                     else {
 194                         System.out.print( " (n/a)" );
 195                     }
 196                     System.out.print( " [" + stats.getN() + "]" );
 197                     System.out.print( " | " );
 198                     System.out.print( ForesterUtil.round( shared_domains_based_genome_distance, 2 ) );
 199                     System.out.print( " | " );
 200                     System.out.println( ForesterUtil.round( shared_binary_combinations_based_genome_distance, 2 ) );
 201                 }
 202                 String pairwise_similarities_output_file_str = automated_pairwise_comparison_prefix + species_i + "_"
 203                         + species_j + automated_pairwise_comparison_suffix;
 204                 switch ( domain_similarity_print_option ) {
 205                     case HTML:
 206                         if ( !pairwise_similarities_output_file_str.endsWith( ".html" ) ) {
 207                             pairwise_similarities_output_file_str += ".html";
 208                         }
 209                         break;
 210                 }
 211                 DescriptiveStatistics pw_stats = null;
 212                 if ( write_pairwise_comparisons ) {
 213                     try {
 214                         final Writer writer = new BufferedWriter( new FileWriter( out_dir == null ? pairwise_similarities_output_file_str
 215                                 : out_dir + ForesterUtil.FILE_SEPARATOR + pairwise_similarities_output_file_str ) );
 216                         pw_stats = SurfacingUtil.writeDomainSimilaritiesToFile( html_desc,
 217                                                                                 new StringBuilder( species_i + "-"
 218                                                                                         + species_j ),
 219                                                                                 writer,
 220                                                                                 null,
 221                                                                                 similarities,
 222                                                                                 true,
 223                                                                                 null,
 224                                                                                 domain_similarity_print_option,
 225                                                                                 domain_similarity_sort_field,
 226                                                                                 scoring,
 227                                                                                 false,
 228                                                                                 tax_code_to_id_map );
 229                     }
 230                     catch ( final IOException e ) {
 231                         ForesterUtil.fatalError( command_line_prg_name, "Failed to write similarites to: \""
 232                                 + pairwise_similarities_output_file_str + "\" [" + e.getMessage() + "]" );
 233                     }
 234                 }
 235                 if ( pw_stats != null ) {
 236                     if ( pw_stats.getMin() >= pw_stats.getMax() ) {
 237                         ForesterUtil
 238                                 .printWarningMessage( command_line_prg_name, "for [" + species_i + "-" + species_j
 239                                         + "] score minimum is [" + pw_stats.getMin() + "] while score maximum is ["
 240                                         + pw_stats.getMax()
 241                                         + "], possibly indicating that a genome is compared to itself" );
 242                     }
 243                 }
 244             }
 245         }
 246         getDomainDistanceScoresMeans().add( domain_distance_scores_means );
 247         getSharedDomainsBasedDistances().add( shared_domains_based_distances );
 248         getSharedBinaryCombinationsBasedDistances().add( shared_binary_combinations_based_distances );
 249         if ( verbose ) {
 250             System.out.println();
 251         }
 252     }
 253
 254     public void performPairwiseComparisonsJacknifed( final Species[] species,
 255                                                      final int number_of_genomes,
 256                                                      final List<GenomeWideCombinableDomains> list_of_genome_wide_combinable_domains,
 257                                                      final boolean verbose,
 258                                                      final int number_of_resamplings,
 259                                                      final double jacknife_ratio,
 260                                                      final long random_seed ) {
 261         init();
 262         if ( number_of_resamplings < 2 ) {
 263             throw new IllegalArgumentException( "attempt to perform jacknife resampling with less than 2 resamplings" );
 264         }
 265         if ( jacknife_ratio <= 0.0 ) {
 266             throw new IllegalArgumentException( "attempt to perform jacknife resampling with jacknife ratio of 0.0 or less" );
 267         }
 268         else if ( jacknife_ratio >= 1.0 ) {
 269             throw new IllegalArgumentException( "attempt to perform jacknife resampling with jacknife ratio 1.0 or more" );
 270         }
 271         final DomainId[] all_unique_domain_ids = getAllUniqueDomainIdAsArray( list_of_genome_wide_combinable_domains );
 272         if ( verbose ) {
 273             System.out.println();
 274             System.out.println( "Jacknife: total of domains: " + all_unique_domain_ids.length );
 275         }
 276         if ( verbose ) {
 277             System.out.print( "resampling " );
 278         }
 279         final Random generator = new Random( random_seed );
 280         for( int r = 0; r < number_of_resamplings; ++r ) {
 281             if ( verbose ) {
 282                 System.out.print( " " + r );
 283             }
 284             final SortedSet<DomainId> domain_ids_to_ignore = randomlyPickDomainIds( all_unique_domain_ids,
 285                                                                                     jacknife_ratio,
 286                                                                                     generator );
 287             final BasicSymmetricalDistanceMatrix shared_domains_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 288             final BasicSymmetricalDistanceMatrix shared_binary_combinations_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 289             for( int i = 0; i < number_of_genomes; ++i ) {
 290                 final String species_i = species[ i ].getSpeciesId();
 291                 shared_domains_based_distances.setIdentifier( i, species_i );
 292                 shared_binary_combinations_based_distances.setIdentifier( i, species_i );
 293                 for( int j = 0; j < i; ++j ) {
 294                     final List<GenomeWideCombinableDomains> genome_pair = new ArrayList<GenomeWideCombinableDomains>( 2 );
 295                     genome_pair.add( list_of_genome_wide_combinable_domains.get( i ) );
 296                     genome_pair.add( list_of_genome_wide_combinable_domains.get( j ) );
 297                     final DomainArchitectureBasedGenomeSimilarityCalculator genome_simiarity_calculator = new DomainArchitectureBasedGenomeSimilarityCalculator( list_of_genome_wide_combinable_domains
 298                                                                                                                                                                          .get( i ),
 299                                                                                                                                                                  list_of_genome_wide_combinable_domains
 300                                                                                                                                                                          .get( j ) );
 301                     genome_simiarity_calculator.setAllowDomainsToBeIgnored( true );
 302                     genome_simiarity_calculator.setDomainIdsToIgnore( domain_ids_to_ignore );
 303                     shared_domains_based_distances.setValue( i, j, 1.0 - genome_simiarity_calculator
 304                             .calculateSharedDomainsBasedGenomeSimilarityScore() );
 305                     shared_binary_combinations_based_distances.setValue( i, j, 1.0 - genome_simiarity_calculator
 306                             .calculateSharedBinaryDomainCombinationBasedGenomeSimilarityScore() );
 307                 }
 308             }
 309             getSharedDomainsBasedDistances().add( shared_domains_based_distances );
 310             getSharedBinaryCombinationsBasedDistances().add( shared_binary_combinations_based_distances );
 311         }
 312         if ( verbose ) {
 313             System.out.println();
 314         }
 315     }
 316
 317     static private DomainId[] getAllUniqueDomainIdAsArray( final List<GenomeWideCombinableDomains> list_of_genome_wide_combinable_domains ) {
 318         DomainId[] all_domain_ids_array;
 319         final SortedSet<DomainId> all_domain_ids = new TreeSet<DomainId>();
 320         for( final GenomeWideCombinableDomains genome_wide_combinable_domains : list_of_genome_wide_combinable_domains ) {
 321             final SortedSet<DomainId> all_domains = genome_wide_combinable_domains.getAllDomainIds();
 322             for( final DomainId domain : all_domains ) {
 323                 all_domain_ids.add( domain );
 324             }
 325         }
 326         all_domain_ids_array = new DomainId[ all_domain_ids.size() ];
 327         int n = 0;
 328         for( final DomainId domain_id : all_domain_ids ) {
 329             all_domain_ids_array[ n++ ] = domain_id;
 330         }
 331         return all_domain_ids_array;
 332     }
 333
 334     static private SortedSet<DomainId> randomlyPickDomainIds( final DomainId[] all_domain_ids_array,
 335                                                               final double jacknife_ratio,
 336                                                               final Random generator ) {
 337         final int size = all_domain_ids_array.length;
 338         final SortedSet<DomainId> random_domain_ids = new TreeSet<DomainId>();
 339         final int number_of_ids_pick = ForesterUtil.roundToInt( jacknife_ratio * size );
 340         while ( random_domain_ids.size() < number_of_ids_pick ) {
 341             final int r = generator.nextInt( size );
 342             random_domain_ids.add( all_domain_ids_array[ r ] );
 343         }
 344         return random_domain_ids;
 345     }
 346 }