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: www.phylosoft.org/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.surfacing.DomainSimilarityCalculator.Detailedness;
  47 import org.forester.util.DescriptiveStatistics;
  48 import org.forester.util.ForesterUtil;
  49
  50 public class PairwiseGenomeComparator {
  51
  52     private List<DistanceMatrix> _domain_distance_scores_means;
  53     private List<DistanceMatrix> _shared_domains_based_distances;
  54     private List<DistanceMatrix> _shared_binary_combinations_based_distances;
  55
  56     //private List<HistogramData>  _histogram_datas;
  57     public PairwiseGenomeComparator() {
  58         init();
  59     }
  60
  61     public List<DistanceMatrix> getDomainDistanceScoresMeans() {
  62         return _domain_distance_scores_means;
  63     }
  64
  65     //public List<HistogramData> getHistogramDatas() {
  66     //    return _histogram_datas;
  67     //}
  68     public List<DistanceMatrix> getSharedBinaryCombinationsBasedDistances() {
  69         return _shared_binary_combinations_based_distances;
  70     }
  71
  72     public List<DistanceMatrix> getSharedDomainsBasedDistances() {
  73         return _shared_domains_based_distances;
  74     }
  75
  76     private void init() {
  77         //_histogram_datas = new ArrayList<HistogramData>();
  78         _domain_distance_scores_means = new ArrayList<DistanceMatrix>();
  79         _shared_domains_based_distances = new ArrayList<DistanceMatrix>();
  80         _shared_binary_combinations_based_distances = new ArrayList<DistanceMatrix>();
  81     }
  82
  83     public void performPairwiseComparisons( final StringBuilder html_desc,
  84                                             final boolean sort_by_species_count_first,
  85                                             final Detailedness detailedness,
  86                                             final boolean ignore_domains_without_combs_in_all_spec,
  87                                             final boolean ignore_domains_specific_to_one_species,
  88                                             final DomainSimilarity.DomainSimilaritySortField domain_similarity_sort_field,
  89                                             final PrintableDomainSimilarity.PRINT_OPTION domain_similarity_print_option,
  90                                             final DomainSimilarity.DomainSimilarityScoring scoring,
  91                                             final Map<DomainId, List<GoId>> domain_id_to_go_ids_map,
  92                                             final Map<GoId, GoTerm> go_id_to_term_map,
  93                                             final GoNameSpace go_namespace_limit,
  94                                             final Species[] species,
  95                                             final int number_of_genomes,
  96                                             final List<GenomeWideCombinableDomains> list_of_genome_wide_combinable_domains,
  97                                             final PairwiseDomainSimilarityCalculator pw_calc,
  98                                             final String automated_pairwise_comparison_suffix,
  99                                             final boolean verbose,
 100                                             final String automated_pairwise_comparison_prefix,
 101                                             final String command_line_prg_name,
 102                                             final File out_dir,
 103                                             final boolean write_pairwise_comparisons ) {
 104         init();
 105         final BasicSymmetricalDistanceMatrix domain_distance_scores_means = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 106         final BasicSymmetricalDistanceMatrix shared_domains_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 107         final BasicSymmetricalDistanceMatrix shared_binary_combinations_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 108         if ( verbose ) {
 109             System.out.println();
 110             System.out.println( "Pairwise genome distances:" );
 111             System.out.print( "[species-i - species-j:" );
 112             System.out.print( " mean-score-based" );
 113             System.out.print( " (sd)" );
 114             System.out.print( " [N]" );
 115             System.out.print( " | shared-domains-based" );
 116             System.out.println( " | shared-binary-combinations-based]" );
 117             System.out.println();
 118         }
 119         for( int i = 0; i < number_of_genomes; ++i ) {
 120             final String species_i = species[ i ].getSpeciesId();
 121             domain_distance_scores_means.setIdentifier( i, species_i );
 122             shared_domains_based_distances.setIdentifier( i, species_i );
 123             shared_binary_combinations_based_distances.setIdentifier( i, species_i );
 124             if ( verbose ) {
 125                 System.out.println( ( i + 1 ) + "/" + number_of_genomes );
 126             }
 127             for( int j = 0; j < i; ++j ) {
 128                 if ( ( list_of_genome_wide_combinable_domains.get( i ).getSize() < 1 )
 129                         || ( list_of_genome_wide_combinable_domains.get( j ).getSize() < 1 ) ) {
 130                     domain_distance_scores_means
 131                             .setValue( i, j, DomainArchitectureBasedGenomeSimilarityCalculator.MAX_SIMILARITY_SCORE );
 132                     shared_domains_based_distances
 133                             .setValue( i, j, DomainArchitectureBasedGenomeSimilarityCalculator.MAX_SIMILARITY_SCORE );
 134                     shared_binary_combinations_based_distances
 135                             .setValue( i, j, DomainArchitectureBasedGenomeSimilarityCalculator.MAX_SIMILARITY_SCORE );
 136                     continue;
 137                 }
 138                 final List<GenomeWideCombinableDomains> genome_pair = new ArrayList<GenomeWideCombinableDomains>( 2 );
 139                 genome_pair.add( list_of_genome_wide_combinable_domains.get( i ) );
 140                 genome_pair.add( list_of_genome_wide_combinable_domains.get( j ) );
 141                 DomainSimilarityCalculator.GoAnnotationOutput go_annotation_output = DomainSimilarityCalculator.GoAnnotationOutput.NONE;
 142                 if ( domain_id_to_go_ids_map != null ) {
 143                     go_annotation_output = DomainSimilarityCalculator.GoAnnotationOutput.ALL;
 144                 }
 145                 final DomainSimilarityCalculator calc = new BasicDomainSimilarityCalculator( domain_similarity_sort_field,
 146                                                                                              sort_by_species_count_first,
 147                                                                                              true );
 148                 final SortedSet<DomainSimilarity> similarities = calc
 149                         .calculateSimilarities( pw_calc,
 150                                                 genome_pair,
 151                                                 ignore_domains_without_combs_in_all_spec,
 152                                                 ignore_domains_specific_to_one_species );
 153                 SurfacingUtil.decoratePrintableDomainSimilarities( similarities,
 154                                                                    detailedness,
 155                                                                    go_annotation_output,
 156                                                                    go_id_to_term_map,
 157                                                                    go_namespace_limit );
 158                 final DescriptiveStatistics stats = SurfacingUtil
 159                         .calculateDescriptiveStatisticsForMeanValues( similarities );
 160                 final String species_j = species[ j ].getSpeciesId();
 161                 final DomainArchitectureBasedGenomeSimilarityCalculator genome_similarity_calculator = new DomainArchitectureBasedGenomeSimilarityCalculator( list_of_genome_wide_combinable_domains
 162                                                                                                                                                                       .get( i ),
 163                                                                                                                                                               list_of_genome_wide_combinable_domains
 164                                                                                                                                                                       .get( j ) );
 165                 genome_similarity_calculator.setAllowDomainsToBeIgnored( false );
 166                 double dissimilarity_score_mean;
 167                 if ( stats.getN() < 1 ) {
 168                     // No domains in common
 169                     dissimilarity_score_mean = 1.0;
 170                 }
 171                 else {
 172                     dissimilarity_score_mean = 1.0 - stats.arithmeticMean();
 173                 }
 174                 final double shared_domains_based_genome_distance = 1.0 - genome_similarity_calculator
 175                         .calculateSharedDomainsBasedGenomeSimilarityScore();
 176                 final double shared_binary_combinations_based_genome_distance = 1.0 - genome_similarity_calculator
 177                         .calculateSharedBinaryDomainCombinationBasedGenomeSimilarityScore();
 178                 domain_distance_scores_means.setValue( i, j, dissimilarity_score_mean );
 179                 shared_domains_based_distances.setValue( i, j, shared_domains_based_genome_distance );
 180                 shared_binary_combinations_based_distances.setValue( i,
 181                                                                      j,
 182                                                                      shared_binary_combinations_based_genome_distance );
 183                 if ( verbose ) {
 184                     System.out.print( species_i + "-" );
 185                     System.out.print( species_j + ": " );
 186                     System.out.print( ForesterUtil.round( dissimilarity_score_mean, 2 ) );
 187                     if ( stats.getN() > 1 ) {
 188                         System.out.print( " (" + ForesterUtil.round( stats.sampleStandardDeviation(), 2 ) + ")" );
 189                     }
 190                     else {
 191                         System.out.print( " (n/a)" );
 192                     }
 193                     System.out.print( " [" + stats.getN() + "]" );
 194                     System.out.print( " | " );
 195                     System.out.print( ForesterUtil.round( shared_domains_based_genome_distance, 2 ) );
 196                     System.out.print( " | " );
 197                     System.out.println( ForesterUtil.round( shared_binary_combinations_based_genome_distance, 2 ) );
 198                 }
 199                 String pairwise_similarities_output_file_str = automated_pairwise_comparison_prefix + species_i + "_"
 200                         + species_j + automated_pairwise_comparison_suffix;
 201                 switch ( domain_similarity_print_option ) {
 202                     case HTML:
 203                         if ( !pairwise_similarities_output_file_str.endsWith( ".html" ) ) {
 204                             pairwise_similarities_output_file_str += ".html";
 205                         }
 206                         break;
 207                 }
 208                 DescriptiveStatistics pw_stats = null;
 209                 if ( write_pairwise_comparisons ) {
 210                     try {
 211                         final Writer writer = new BufferedWriter( new FileWriter( out_dir == null ? pairwise_similarities_output_file_str
 212                                 : out_dir + ForesterUtil.FILE_SEPARATOR + pairwise_similarities_output_file_str ) );
 213                         pw_stats = SurfacingUtil.writeDomainSimilaritiesToFile( html_desc,
 214                                                                                 new StringBuilder( species_i + "-"
 215                                                                                         + species_j ),
 216                                                                                 writer,
 217                                                                                 similarities,
 218                                                                                 true,
 219                                                                                 null,
 220                                                                                 domain_similarity_print_option,
 221                                                                                 domain_similarity_sort_field,
 222                                                                                 scoring,
 223                                                                                 false );
 224                     }
 225                     catch ( final IOException e ) {
 226                         ForesterUtil.fatalError( command_line_prg_name, "Failed to write similarites to: \""
 227                                 + pairwise_similarities_output_file_str + "\" [" + e.getMessage() + "]" );
 228                     }
 229                 }
 230                 if ( pw_stats != null ) {
 231                     if ( pw_stats.getMin() >= pw_stats.getMax() ) {
 232                         ForesterUtil
 233                                 .printWarningMessage( command_line_prg_name, "for [" + species_i + "-" + species_j
 234                                         + "] score minimum is [" + pw_stats.getMin() + "] while score maximum is ["
 235                                         + pw_stats.getMax()
 236                                         + "], possibly indicating that a genome is compared to itself" );
 237                     }
 238                 }
 239             }
 240         }
 241         getDomainDistanceScoresMeans().add( domain_distance_scores_means );
 242         getSharedDomainsBasedDistances().add( shared_domains_based_distances );
 243         getSharedBinaryCombinationsBasedDistances().add( shared_binary_combinations_based_distances );
 244         if ( verbose ) {
 245             System.out.println();
 246         }
 247     }
 248
 249     public void performPairwiseComparisonsJacknifed( final Species[] species,
 250                                                      final int number_of_genomes,
 251                                                      final List<GenomeWideCombinableDomains> list_of_genome_wide_combinable_domains,
 252                                                      final boolean verbose,
 253                                                      final int number_of_resamplings,
 254                                                      final double jacknife_ratio,
 255                                                      final long random_seed ) {
 256         init();
 257         if ( number_of_resamplings < 2 ) {
 258             throw new IllegalArgumentException( "attempt to perform jacknife resampling with less than 2 resamplings" );
 259         }
 260         if ( jacknife_ratio <= 0.0 ) {
 261             throw new IllegalArgumentException( "attempt to perform jacknife resampling with jacknife ratio of 0.0 or less" );
 262         }
 263         else if ( jacknife_ratio >= 1.0 ) {
 264             throw new IllegalArgumentException( "attempt to perform jacknife resampling with jacknife ratio 1.0 or more" );
 265         }
 266         final DomainId[] all_unique_domain_ids = getAllUniqueDomainIdAsArray( list_of_genome_wide_combinable_domains );
 267         if ( verbose ) {
 268             System.out.println();
 269             System.out.println( "Jacknife: total of domains: " + all_unique_domain_ids.length );
 270         }
 271         if ( verbose ) {
 272             System.out.print( "resampling " );
 273         }
 274         final Random generator = new Random( random_seed );
 275         for( int r = 0; r < number_of_resamplings; ++r ) {
 276             if ( verbose ) {
 277                 System.out.print( " " + r );
 278             }
 279             final SortedSet<DomainId> domain_ids_to_ignore = randomlyPickDomainIds( all_unique_domain_ids,
 280                                                                                     jacknife_ratio,
 281                                                                                     generator );
 282             final BasicSymmetricalDistanceMatrix shared_domains_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 283             final BasicSymmetricalDistanceMatrix shared_binary_combinations_based_distances = new BasicSymmetricalDistanceMatrix( number_of_genomes );
 284             for( int i = 0; i < number_of_genomes; ++i ) {
 285                 final String species_i = species[ i ].getSpeciesId();
 286                 shared_domains_based_distances.setIdentifier( i, species_i );
 287                 shared_binary_combinations_based_distances.setIdentifier( i, species_i );
 288                 for( int j = 0; j < i; ++j ) {
 289                     final List<GenomeWideCombinableDomains> genome_pair = new ArrayList<GenomeWideCombinableDomains>( 2 );
 290                     genome_pair.add( list_of_genome_wide_combinable_domains.get( i ) );
 291                     genome_pair.add( list_of_genome_wide_combinable_domains.get( j ) );
 292                     final DomainArchitectureBasedGenomeSimilarityCalculator genome_simiarity_calculator = new DomainArchitectureBasedGenomeSimilarityCalculator( list_of_genome_wide_combinable_domains
 293                                                                                                                                                                          .get( i ),
 294                                                                                                                                                                  list_of_genome_wide_combinable_domains
 295                                                                                                                                                                          .get( j ) );
 296                     genome_simiarity_calculator.setAllowDomainsToBeIgnored( true );
 297                     genome_simiarity_calculator.setDomainIdsToIgnore( domain_ids_to_ignore );
 298                     shared_domains_based_distances.setValue( i, j, 1.0 - genome_simiarity_calculator
 299                             .calculateSharedDomainsBasedGenomeSimilarityScore() );
 300                     shared_binary_combinations_based_distances.setValue( i, j, 1.0 - genome_simiarity_calculator
 301                             .calculateSharedBinaryDomainCombinationBasedGenomeSimilarityScore() );
 302                 }
 303             }
 304             getSharedDomainsBasedDistances().add( shared_domains_based_distances );
 305             getSharedBinaryCombinationsBasedDistances().add( shared_binary_combinations_based_distances );
 306         }
 307         if ( verbose ) {
 308             System.out.println();
 309         }
 310     }
 311
 312     static private DomainId[] getAllUniqueDomainIdAsArray( final List<GenomeWideCombinableDomains> list_of_genome_wide_combinable_domains ) {
 313         DomainId[] all_domain_ids_array;
 314         final SortedSet<DomainId> all_domain_ids = new TreeSet<DomainId>();
 315         for( final GenomeWideCombinableDomains genome_wide_combinable_domains : list_of_genome_wide_combinable_domains ) {
 316             final SortedSet<DomainId> all_domains = genome_wide_combinable_domains.getAllDomainIds();
 317             for( final DomainId domain : all_domains ) {
 318                 all_domain_ids.add( domain );
 319             }
 320         }
 321         all_domain_ids_array = new DomainId[ all_domain_ids.size() ];
 322         int n = 0;
 323         for( final DomainId domain_id : all_domain_ids ) {
 324             all_domain_ids_array[ n++ ] = domain_id;
 325         }
 326         return all_domain_ids_array;
 327     }
 328
 329     static private SortedSet<DomainId> randomlyPickDomainIds( final DomainId[] all_domain_ids_array,
 330                                                               final double jacknife_ratio,
 331                                                               final Random generator ) {
 332         final int size = all_domain_ids_array.length;
 333         final SortedSet<DomainId> random_domain_ids = new TreeSet<DomainId>();
 334         final int number_of_ids_pick = ForesterUtil.roundToInt( jacknife_ratio * size );
 335         while ( random_domain_ids.size() < number_of_ids_pick ) {
 336             final int r = generator.nextInt( size );
 337             random_domain_ids.add( all_domain_ids_array[ r ] );
 338         }
 339         return random_domain_ids;
 340     }
 341 }