forester/java/src/org/forester/tools/PhylogenyDecorator.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: https://sites.google.com/site/cmzmasek/home/software/forester
  25
  26 package org.forester.tools;
  27
  28 import java.io.File;
  29 import java.io.IOException;
  30 import java.util.HashMap;
  31 import java.util.Map;
  32 import java.util.regex.Matcher;
  33
  34 import org.forester.io.parsers.nhx.NHXFormatException;
  35 import org.forester.io.parsers.phyloxml.PhyloXmlDataFormatException;
  36 import org.forester.io.parsers.util.ParserUtils;
  37 import org.forester.phylogeny.Phylogeny;
  38 import org.forester.phylogeny.PhylogenyNode;
  39 import org.forester.phylogeny.data.Accession;
  40 import org.forester.phylogeny.data.Annotation;
  41 import org.forester.phylogeny.data.DomainArchitecture;
  42 import org.forester.phylogeny.data.Identifier;
  43 import org.forester.phylogeny.data.Sequence;
  44 import org.forester.phylogeny.iterators.PhylogenyNodeIterator;
  45 import org.forester.util.BasicTable;
  46 import org.forester.util.BasicTableParser;
  47 import org.forester.util.ForesterUtil;
  48
  49 public final class PhylogenyDecorator {
  50
  51     final private static String TP_NODE_NAME            = "NODE_NAME";
  52     final private static String TP_SEQ_ACCESSION        = "SEQ_ACCESSION";
  53     final private static String TP_SEQ_ACCESSION_SOURCE = "SEQ_ACCESSION_SOURCE";
  54     final private static String TP_SEQ_ANNOTATION_DESC  = "SEQ_ANNOTATION_DESC";
  55     final private static String TP_SEQ_ANNOTATION_REF   = "SEQ_ANNOTATION_REF";
  56     final private static String TP_SEQ_MOL_SEQ          = "SEQ_MOL_SEQ";
  57     final private static String TP_SEQ_NAME             = "SEQ_NAME";
  58     final private static String TP_SEQ_SYMBOL           = "SEQ_SYMBOL";
  59     final private static String TP_TAXONOMY_CN          = "TAXONOMY_CN";
  60     // From evoruby/lib/evo/apps/tseq_taxonomy_processor.rb:
  61     final private static String TP_TAXONOMY_CODE        = "TAXONOMY_CODE";
  62     final private static String TP_TAXONOMY_ID          = "TAXONOMY_ID";
  63     final private static String TP_TAXONOMY_ID_PROVIDER = "TAXONOMY_ID_PROVIDER";
  64     final private static String TP_TAXONOMY_SN          = "TAXONOMY_SN";
  65     final private static String TP_TAXONOMY_SYN         = "TAXONOMY_SYN";
  66
  67     private PhylogenyDecorator() {
  68         // Not needed.
  69     }
  70
  71     public static void decorate( final Phylogeny phylogeny,
  72                                  final Map<String, Map<String, String>> map,
  73                                  final boolean picky ) throws IllegalArgumentException, PhyloXmlDataFormatException {
  74         for( final PhylogenyNodeIterator iter = phylogeny.iteratorPostorder(); iter.hasNext(); ) {
  75             final PhylogenyNode node = iter.next();
  76             final String name = node.getName();
  77             if ( !ForesterUtil.isEmpty( name ) ) {
  78                 if ( map.containsKey( name ) ) {
  79                     final Map<String, String> new_values = map.get( name );
  80                     if ( new_values != null ) {
  81                         if ( new_values.containsKey( TP_TAXONOMY_CODE ) ) {
  82                             ForesterUtil.ensurePresenceOfTaxonomy( node );
  83                             node.getNodeData().getTaxonomy().setTaxonomyCode( new_values.get( TP_TAXONOMY_CODE ) );
  84                         }
  85                         if ( new_values.containsKey( TP_TAXONOMY_ID )
  86                                 && new_values.containsKey( TP_TAXONOMY_ID_PROVIDER ) ) {
  87                             ForesterUtil.ensurePresenceOfTaxonomy( node );
  88                             node.getNodeData()
  89                                     .getTaxonomy()
  90                                     .setIdentifier( new Identifier( new_values.get( TP_TAXONOMY_ID ),
  91                                                                     new_values.get( TP_TAXONOMY_ID_PROVIDER ) ) );
  92                         }
  93                         else if ( new_values.containsKey( TP_TAXONOMY_ID ) ) {
  94                             ForesterUtil.ensurePresenceOfTaxonomy( node );
  95                             node.getNodeData().getTaxonomy()
  96                                     .setIdentifier( new Identifier( new_values.get( TP_TAXONOMY_ID ) ) );
  97                         }
  98                         if ( new_values.containsKey( TP_TAXONOMY_SN ) ) {
  99                             ForesterUtil.ensurePresenceOfTaxonomy( node );
 100                             node.getNodeData().getTaxonomy().setScientificName( new_values.get( TP_TAXONOMY_SN ) );
 101                         }
 102                         if ( new_values.containsKey( TP_TAXONOMY_CN ) ) {
 103                             ForesterUtil.ensurePresenceOfTaxonomy( node );
 104                             node.getNodeData().getTaxonomy().setCommonName( new_values.get( TP_TAXONOMY_CN ) );
 105                         }
 106                         if ( new_values.containsKey( TP_TAXONOMY_SYN ) ) {
 107                             ForesterUtil.ensurePresenceOfTaxonomy( node );
 108                             node.getNodeData().getTaxonomy().getSynonyms().add( new_values.get( TP_TAXONOMY_SYN ) );
 109                         }
 110                         if ( new_values.containsKey( TP_SEQ_ACCESSION )
 111                                 && new_values.containsKey( TP_SEQ_ACCESSION_SOURCE ) ) {
 112                             ForesterUtil.ensurePresenceOfSequence( node );
 113                             node.getNodeData()
 114                                     .getSequence()
 115                                     .setAccession( new Accession( new_values.get( TP_SEQ_ACCESSION ),
 116                                                                   new_values.get( TP_SEQ_ACCESSION_SOURCE ) ) );
 117                         }
 118                         if ( new_values.containsKey( TP_SEQ_ANNOTATION_DESC ) ) {
 119                             ForesterUtil.ensurePresenceOfSequence( node );
 120                             final Annotation ann = new Annotation();
 121                             ann.setDesc( new_values.get( TP_SEQ_ANNOTATION_DESC ) );
 122                             node.getNodeData().getSequence().addAnnotation( ann );
 123                         }
 124                         if ( new_values.containsKey( TP_SEQ_ANNOTATION_REF ) ) {
 125                             ForesterUtil.ensurePresenceOfSequence( node );
 126                             final Annotation ann = new Annotation( new_values.get( TP_SEQ_ANNOTATION_REF ) );
 127                             node.getNodeData().getSequence().addAnnotation( ann );
 128                         }
 129                         if ( new_values.containsKey( TP_SEQ_SYMBOL ) ) {
 130                             ForesterUtil.ensurePresenceOfSequence( node );
 131                             node.getNodeData().getSequence().setSymbol( new_values.get( TP_SEQ_SYMBOL ) );
 132                         }
 133                         if ( new_values.containsKey( TP_SEQ_NAME ) ) {
 134                             ForesterUtil.ensurePresenceOfSequence( node );
 135                             node.getNodeData().getSequence().setName( new_values.get( TP_SEQ_NAME ) );
 136                         }
 137                         if ( new_values.containsKey( TP_SEQ_MOL_SEQ ) ) {
 138                             ForesterUtil.ensurePresenceOfSequence( node );
 139                             node.getNodeData().getSequence().setMolecularSequence( new_values.get( TP_SEQ_MOL_SEQ ) );
 140                         }
 141                         if ( new_values.containsKey( TP_NODE_NAME ) ) {
 142                             node.setName( new_values.get( TP_NODE_NAME ) );
 143                         }
 144                     } // if ( new_values != null )
 145                 } // if ( map.containsKey( name ) || ( numbers_of_chars_allowed_to_remove_if_not_found_in_map > 0 ) )
 146                 else if ( picky ) {
 147                     throw new IllegalArgumentException( "\"" + name + "\" not found in name map" );
 148                 }
 149             }
 150         }
 151     }
 152
 153     public static String decorate( final Phylogeny phylogeny,
 154                                    final Map<String, String> map,
 155                                    final FIELD field,
 156                                    final boolean extract_bracketed_scientific_name,
 157                                    final boolean extract_bracketed_tax_code,
 158                                    final boolean picky,
 159                                    final boolean cut_name_after_space,
 160                                    final boolean trim_after_tilde,
 161                                    final boolean verbose ) throws IllegalArgumentException, NHXFormatException,
 162             PhyloXmlDataFormatException {
 163         return PhylogenyDecorator.decorate( phylogeny,
 164                                             map,
 165                                             field,
 166                                             extract_bracketed_scientific_name,
 167                                             extract_bracketed_tax_code,
 168                                             picky,
 169                                             null,
 170                                             cut_name_after_space,
 171                                             trim_after_tilde,
 172                                             verbose );
 173     }
 174
 175     /**
 176      *
 177      *
 178      *
 179      * @param phylogeny
 180      * @param map
 181      *            maps names (in phylogeny) to new values if intermediate_map is
 182      *            null otherwise maps intermediate value to new value
 183      * @param field
 184      * @param picky
 185      * @param intermediate_map
 186      *            maps name (in phylogeny) to a intermediate value
 187      * @throws IllegalArgumentException
 188      * @throws PhyloXmlDataFormatException
 189      */
 190     public static String decorate( final Phylogeny phylogeny,
 191                                    final Map<String, String> map,
 192                                    final FIELD field,
 193                                    final boolean extract_bracketed_scientific_name,
 194                                    final boolean extract_bracketed_tax_code,
 195                                    final boolean picky,
 196                                    final Map<String, String> intermediate_map,
 197                                    final boolean cut_name_after_space,
 198                                    final boolean trim_after_tilde,
 199                                    final boolean verbose ) throws IllegalArgumentException, PhyloXmlDataFormatException {
 200         if ( extract_bracketed_scientific_name && ( field == FIELD.TAXONOMY_SCIENTIFIC_NAME ) ) {
 201             throw new IllegalArgumentException( "attempt to extract bracketed scientific name together with data field pointing to scientific name" );
 202         }
 203         if ( map.isEmpty() ) {
 204             throw new IllegalArgumentException( "map is empty" );
 205         }
 206         if ( picky && ( map.size() < phylogeny.getNumberOfExternalNodes() ) ) {
 207             throw new IllegalArgumentException( "map contains less entries than the tree has external nodes" );
 208         }
 209         int ext_nodes = 0;
 210         int ext_nodes_updated = 0;
 211         int int_nodes = 0;
 212         int int_nodes_updated = 0;
 213         for( final PhylogenyNodeIterator iter = phylogeny.iteratorPostorder(); iter.hasNext(); ) {
 214             final PhylogenyNode node = iter.next();
 215             if ( node.isExternal() ) {
 216                 ++ext_nodes;
 217             }
 218             else {
 219                 ++int_nodes;
 220             }
 221             String name = node.getName();
 222             if ( picky && node.isExternal() && ForesterUtil.isEmpty( name ) ) {
 223                 throw new IllegalArgumentException( "external node with no name present" );
 224             }
 225             String tilde_annotation = null;
 226             if ( trim_after_tilde && ( name.indexOf( '~' ) > 0 ) ) {
 227                 final int ti = name.indexOf( '~' );
 228                 final String orig = name;
 229                 tilde_annotation = name.substring( ti );
 230                 name = name.substring( 0, ti );
 231                 if ( node.isExternal() && ForesterUtil.isEmpty( name ) ) {
 232                     throw new IllegalArgumentException( "external node with illegal name: " + orig );
 233                 }
 234             }
 235             if ( !ForesterUtil.isEmpty( name ) ) {
 236                 if ( intermediate_map != null ) {
 237                     name = PhylogenyDecorator.extractIntermediate( intermediate_map, name, verbose );
 238                 }
 239                 if ( map.containsKey( name ) ) {
 240                     String new_value = map.get( name ).trim().replaceAll( "/\\s+/", " " );
 241                     if ( !ForesterUtil.isEmpty( new_value ) ) {
 242                         if ( node.isExternal() ) {
 243                             ++ext_nodes_updated;
 244                         }
 245                         else {
 246                             ++int_nodes_updated;
 247                         }
 248                         if ( extract_bracketed_scientific_name && new_value.endsWith( "]" ) ) {
 249                             new_value = extractBracketedScientificNames( node, new_value );
 250                         }
 251                         else if ( extract_bracketed_tax_code ) {
 252                             if ( ParserUtils.TAXOMONY_CODE_PATTERN_BRACKETED.matcher( new_value ).find() ) {
 253                                 new_value = extractBracketedTaxCodes( node, new_value );
 254                             }
 255                             else if ( picky ) {
 256                                 throw new IllegalArgumentException( " could not get taxonomy from \"" + new_value
 257                                         + "\"" );
 258                             }
 259                         }
 260                         switch ( field ) {
 261                             case MOL_SEQ:
 262                                 if ( verbose ) {
 263                                     System.out.println( name + ": " + new_value );
 264                                 }
 265                                 if ( !node.getNodeData().isHasSequence() ) {
 266                                     node.getNodeData().setSequence( new Sequence() );
 267                                 }
 268                                 node.getNodeData().getSequence().setMolecularSequence( new_value );
 269                                 break;
 270                             case SEQUENCE_ANNOTATION_DESC:
 271                                 if ( verbose ) {
 272                                     System.out.println( name + ": " + new_value );
 273                                 }
 274                                 if ( !node.getNodeData().isHasSequence() ) {
 275                                     node.getNodeData().setSequence( new Sequence() );
 276                                 }
 277                                 final Annotation annotation = new Annotation();
 278                                 annotation.setDesc( new_value );
 279                                 node.getNodeData().getSequence().addAnnotation( annotation );
 280                                 break;
 281                             case DOMAIN_STRUCTURE:
 282                                 if ( verbose ) {
 283                                     System.out.println( name + ": " + new_value );
 284                                 }
 285                                 if ( !node.getNodeData().isHasSequence() ) {
 286                                     node.getNodeData().setSequence( new Sequence() );
 287                                 }
 288                                 node.getNodeData().getSequence()
 289                                         .setDomainArchitecture( new DomainArchitecture( new_value ) );
 290                                 break;
 291                             case TAXONOMY_CODE:
 292                                 if ( verbose ) {
 293                                     System.out.println( name + ": " + new_value );
 294                                 }
 295                                 ForesterUtil.ensurePresenceOfTaxonomy( node );
 296                                 node.getNodeData().getTaxonomy().setTaxonomyCode( new_value );
 297                                 break;
 298                             case TAXONOMY_SCIENTIFIC_NAME:
 299                                 if ( verbose ) {
 300                                     System.out.println( name + ": " + new_value );
 301                                 }
 302                                 ForesterUtil.ensurePresenceOfTaxonomy( node );
 303                                 node.getNodeData().getTaxonomy().setScientificName( new_value );
 304                                 break;
 305                             case SEQUENCE_NAME:
 306                                 if ( trim_after_tilde ) {
 307                                     new_value = addTildeAnnotation( tilde_annotation, new_value );
 308                                 }
 309                                 if ( verbose ) {
 310                                     System.out.println( name + ": " + new_value );
 311                                 }
 312                                 if ( !node.getNodeData().isHasSequence() ) {
 313                                     node.getNodeData().setSequence( new Sequence() );
 314                                 }
 315                                 node.getNodeData().getSequence().setName( new_value );
 316                                 break;
 317                             case NODE_NAME:
 318                                 if ( verbose ) {
 319                                     System.out.print( name + " -> " );
 320                                 }
 321                                 if ( cut_name_after_space ) {
 322                                     if ( verbose ) {
 323                                         System.out.print( new_value + " -> " );
 324                                     }
 325                                     new_value = PhylogenyDecorator.deleteAtFirstSpace( new_value );
 326                                 }
 327                                 if ( trim_after_tilde ) {
 328                                     new_value = addTildeAnnotation( tilde_annotation, new_value );
 329                                 }
 330                                 if ( verbose ) {
 331                                     System.out.println( new_value );
 332                                 }
 333                                 node.setName( new_value );
 334                                 break;
 335                             default:
 336                                 throw new RuntimeException( "unknown field \"" + field + "\"" );
 337                         }
 338                     }
 339                     else {
 340                         throw new IllegalArgumentException( "node name \"" + name + "\" maps to empty value" );
 341                     }
 342                 }
 343                 else if ( picky ) {
 344                     throw new IllegalArgumentException( "node name \"" + name + "\" not found in map" );
 345                 }
 346             }
 347         }
 348         return "updated " + ext_nodes_updated + "/" + ext_nodes + " external nodes, updated " + int_nodes_updated + "/"
 349                 + int_nodes + " internal nodes";
 350     }
 351
 352     public static Map<String, Map<String, String>> parseMappingTable( final File mapping_table_file )
 353             throws IOException {
 354         final Map<String, Map<String, String>> map = new HashMap<String, Map<String, String>>();
 355         BasicTable<String> mapping_table = null;
 356         mapping_table = BasicTableParser.parse( mapping_table_file, '\t', false, false );
 357         for( int row = 0; row < mapping_table.getNumberOfRows(); ++row ) {
 358             final Map<String, String> row_map = new HashMap<String, String>();
 359             String name = null;
 360             for( int col = 0; col < mapping_table.getNumberOfColumns(); ++col ) {
 361                 final String table_cell = mapping_table.getValue( col, row );
 362                 if ( col == 0 ) {
 363                     name = table_cell;
 364                 }
 365                 else if ( table_cell != null ) {
 366                     final String key = table_cell.substring( 0, table_cell.indexOf( ':' ) );
 367                     final String val = table_cell.substring( table_cell.indexOf( ':' ) + 1, table_cell.length() );
 368                     row_map.put( key, val );
 369                 }
 370             }
 371             map.put( name, row_map );
 372         }
 373         return map;
 374     }
 375
 376     private final static String addTildeAnnotation( final String tilde_annotation, final String new_value ) {
 377         if ( ForesterUtil.isEmpty( tilde_annotation ) ) {
 378             return new_value;
 379         }
 380         return new_value + tilde_annotation;
 381     }
 382
 383     private static String deleteAtFirstSpace( final String name ) {
 384         final int first_space = name.indexOf( " " );
 385         if ( first_space > 1 ) {
 386             return name.substring( 0, first_space ).trim();
 387         }
 388         return name;
 389     }
 390
 391     private static String extractBracketedScientificNames( final PhylogenyNode node, final String new_value ) {
 392         final int i = new_value.lastIndexOf( "[" );
 393         final String scientific_name = new_value.substring( i + 1, new_value.length() - 1 );
 394         ForesterUtil.ensurePresenceOfTaxonomy( node );
 395         node.getNodeData().getTaxonomy().setScientificName( scientific_name );
 396         return new_value.substring( 0, i - 1 ).trim();
 397     }
 398
 399     private static String extractBracketedTaxCodes( final PhylogenyNode node, final String new_value ) {
 400         final StringBuilder sb = new StringBuilder();
 401         sb.append( new_value );
 402         final String tc = extractBracketedTaxCodes( sb );
 403         if ( !ForesterUtil.isEmpty( tc ) ) {
 404             ForesterUtil.ensurePresenceOfTaxonomy( node );
 405             try {
 406                 node.getNodeData().getTaxonomy().setTaxonomyCode( tc );
 407             }
 408             catch ( final PhyloXmlDataFormatException e ) {
 409                 throw new IllegalArgumentException( "illegal format for taxonomy code: " + tc );
 410             }
 411             return sb.toString().trim();
 412         }
 413         return new_value;
 414     }
 415
 416     private static String extractBracketedTaxCodes( final StringBuilder sb ) {
 417         final Matcher m = ParserUtils.TAXOMONY_CODE_PATTERN_BRACKETED.matcher( sb );
 418         if ( m.find() ) {
 419             final String tc = m.group( 1 );
 420             sb.delete( m.start( 1 ) - 1, m.end( 1 ) + 1 );
 421             return tc;
 422         }
 423         return null;
 424     }
 425
 426     private static String extractIntermediate( final Map<String, String> intermediate_map,
 427                                                final String name,
 428                                                final boolean verbose ) {
 429         String new_name = null;
 430         if ( verbose ) {
 431             System.out.print( name + " => " );
 432         }
 433         if ( intermediate_map.containsKey( name ) ) {
 434             new_name = intermediate_map.get( name );
 435             if ( ForesterUtil.isEmpty( new_name ) ) {
 436                 throw new IllegalArgumentException( "\"" + name + "\" maps to null or empty string in secondary map" );
 437             }
 438         }
 439         else {
 440             throw new IllegalArgumentException( "\"" + name + "\" not found in name secondary map" );
 441         }
 442         if ( verbose ) {
 443             System.out.println( new_name + "  " );
 444         }
 445         return new_name;
 446     }
 447
 448     public static enum FIELD {
 449         DOMAIN_STRUCTURE,
 450         MOL_SEQ,
 451         NODE_NAME,
 452         SEQUENCE_ANNOTATION_DESC,
 453         SEQUENCE_NAME,
 454         TAXONOMY_CODE,
 455         TAXONOMY_SCIENTIFIC_NAME;
 456     }
 457 }