in progress...
[jalview.git] / forester / java / src / org / forester / evoinference / parsimony / SankoffParsimony.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
26
27 package org.forester.evoinference.parsimony;
28
29 import java.util.ArrayList;
30 import java.util.HashMap;
31 import java.util.Iterator;
32 import java.util.List;
33 import java.util.Map;
34 import java.util.Random;
35 import java.util.SortedSet;
36 import java.util.TreeSet;
37
38 import org.forester.evoinference.matrix.character.BasicCharacterStateMatrix;
39 import org.forester.evoinference.matrix.character.CharacterStateMatrix;
40 import org.forester.evoinference.matrix.character.CharacterStateMatrix.BinaryStates;
41 import org.forester.evoinference.matrix.character.CharacterStateMatrix.GainLossStates;
42 import org.forester.phylogeny.Phylogeny;
43 import org.forester.phylogeny.PhylogenyNode;
44 import org.forester.phylogeny.iterators.PhylogenyNodeIterator;
45 import org.forester.util.ForesterUtil;
46
47 /**
48  *
49  * IN PROGRESS!
50  * DO NOT USE!
51  *
52  *
53  * @param <STATE_TYPE>
54  */
55 public class SankoffParsimony<STATE_TYPE> {
56
57     final static private BinaryStates              PRESENT                         = BinaryStates.PRESENT;
58     final static private BinaryStates              ABSENT                          = BinaryStates.ABSENT;
59     final static private GainLossStates            LOSS                            = GainLossStates.LOSS;
60     final static private GainLossStates            GAIN                            = GainLossStates.GAIN;
61     final static private GainLossStates            UNCHANGED_PRESENT               = GainLossStates.UNCHANGED_PRESENT;
62     final static private GainLossStates            UNCHANGED_ABSENT                = GainLossStates.UNCHANGED_ABSENT;
63     private static final boolean                   RETURN_INTERNAL_STATES_DEFAULT  = false;
64     private static final boolean                   RETURN_GAIN_LOSS_MATRIX_DEFAULT = false;
65     private static final boolean                   RANDOMIZE_DEFAULT               = false;
66     private static final long                      RANDOM_NUMBER_SEED_DEFAULT      = 21;
67     private static final boolean                   USE_LAST_DEFAULT                = false;
68     private boolean                                _return_internal_states         = false;
69     private boolean                                _return_gain_loss               = false;
70     private int                                    _total_gains;
71     private int                                    _total_losses;
72     private int                                    _total_unchanged;
73     private CharacterStateMatrix<List<STATE_TYPE>> _internal_states_matrix_prior_to_traceback;
74     private CharacterStateMatrix<STATE_TYPE>       _internal_states_matrix_after_traceback;
75     private CharacterStateMatrix<GainLossStates>   _gain_loss_matrix;
76     private boolean                                _randomize;
77     private boolean                                _use_last;
78     private int                                    _cost;
79     private long                                   _random_number_seed;
80     private Random                                 _random_generator;
81
82     public SankoffParsimony() {
83         init();
84     }
85
86     private int determineIndex( final SortedSet<STATE_TYPE> current_node_states, int i ) {
87         if ( isRandomize() ) {
88             i = getRandomGenerator().nextInt( current_node_states.size() );
89         }
90         else if ( isUseLast() ) {
91             i = current_node_states.size() - 1;
92         }
93         return i;
94     }
95
96     public void execute( final Phylogeny p, final CharacterStateMatrix<STATE_TYPE> external_node_states_matrix ) {
97         if ( !p.isRooted() ) {
98             throw new IllegalArgumentException( "attempt to execute Fitch parsimony on unroored phylogeny" );
99         }
100         if ( external_node_states_matrix.isEmpty() ) {
101             throw new IllegalArgumentException( "character matrix is empty" );
102         }
103         if ( external_node_states_matrix.getNumberOfIdentifiers() != p.getNumberOfExternalNodes() ) {
104             throw new IllegalArgumentException( "number of external nodes in phylogeny ["
105                     + p.getNumberOfExternalNodes() + "] and number of indentifiers ["
106                     + external_node_states_matrix.getNumberOfIdentifiers() + "] in matrix are not equal" );
107         }
108         reset();
109         if ( isReturnInternalStates() ) {
110             initializeInternalStates( p, external_node_states_matrix );
111         }
112         if ( isReturnGainLossMatrix() ) {
113             initializeGainLossMatrix( p, external_node_states_matrix );
114         }
115         for( int character_index = 0; character_index < external_node_states_matrix.getNumberOfCharacters(); ++character_index ) {
116             executeForOneCharacter( p,
117                                     getStatesForCharacter( p, external_node_states_matrix, character_index ),
118                                     getStatesForCharacterForTraceback( p, external_node_states_matrix, character_index ),
119                                     character_index );
120         }
121         if ( external_node_states_matrix.getState( 0, 0 ) instanceof BinaryStates ) {
122             if ( ( external_node_states_matrix.getNumberOfCharacters() * p.getNumberOfBranches() ) != ( getTotalGains()
123                     + getTotalLosses() + getTotalUnchanged() ) ) {
124                 throw new RuntimeException( "this should not have happened: something is deeply wrong with Fitch parsimony implementation" );
125             }
126         }
127     }
128
129     private void executeForOneCharacter( final Phylogeny p,
130                                          final Map<PhylogenyNode, SortedSet<STATE_TYPE>> states,
131                                          final Map<PhylogenyNode, STATE_TYPE> traceback_states,
132                                          final int character_state_column ) {
133         postOrderTraversal( p, states );
134         preOrderTraversal( p, states, traceback_states, character_state_column );
135     }
136
137     public int getCost() {
138         return _cost;
139     }
140
141     public CharacterStateMatrix<CharacterStateMatrix.GainLossStates> getGainLossMatrix() {
142         if ( !isReturnGainLossMatrix() ) {
143             throw new RuntimeException( "creation of gain-loss matrix has not been enabled" );
144         }
145         return _gain_loss_matrix;
146     }
147
148     public CharacterStateMatrix<STATE_TYPE> getInternalStatesMatrix() {
149         if ( !isReturnInternalStates() ) {
150             throw new RuntimeException( "creation of internal state matrix has not been enabled" );
151         }
152         return _internal_states_matrix_after_traceback;
153     }
154
155     /**
156      * Returns a view of the internal states prior to trace-back.
157      *
158      * @return
159      */
160     public CharacterStateMatrix<List<STATE_TYPE>> getInternalStatesMatrixPriorToTraceback() {
161         if ( !isReturnInternalStates() ) {
162             throw new RuntimeException( "creation of internal state matrix has not been enabled" );
163         }
164         return _internal_states_matrix_prior_to_traceback;
165     }
166
167     private SortedSet<STATE_TYPE> getIntersectionOfStatesOfChildNodes( final Map<PhylogenyNode, SortedSet<STATE_TYPE>> states,
168                                                                        final PhylogenyNode node ) throws AssertionError {
169         final SortedSet<STATE_TYPE> states_in_child_nodes = new TreeSet<STATE_TYPE>();
170         for( int i = 0; i < node.getNumberOfDescendants(); ++i ) {
171             final PhylogenyNode node_child = node.getChildNode( i );
172             if ( !states.containsKey( node_child ) ) {
173                 throw new AssertionError( "this should not have happened: node [" + node_child.getName()
174                                           + "] not found in node state map" );
175             }
176             if ( i == 0 ) {
177                 states_in_child_nodes.addAll( states.get( node_child ) );
178             }
179             else {
180                 states_in_child_nodes.retainAll( states.get( node_child ) );
181             }
182         }
183         return states_in_child_nodes;
184     }
185
186     private Random getRandomGenerator() {
187         return _random_generator;
188     }
189
190     private long getRandomNumberSeed() {
191         return _random_number_seed;
192     }
193
194     private STATE_TYPE getStateAt( final int i, final SortedSet<STATE_TYPE> states ) {
195         final Iterator<STATE_TYPE> it = states.iterator();
196         for( int j = 0; j < i; ++j ) {
197             it.next();
198         }
199         return it.next();
200     }
201
202     private Map<PhylogenyNode, SortedSet<STATE_TYPE>> getStatesForCharacter( final Phylogeny p,
203                                                                              final CharacterStateMatrix<STATE_TYPE> matrix,
204                                                                              final int character_index ) {
205         final Map<PhylogenyNode, SortedSet<STATE_TYPE>> states = new HashMap<PhylogenyNode, SortedSet<STATE_TYPE>>( matrix
206                 .getNumberOfIdentifiers() );
207         for( int indentifier_index = 0; indentifier_index < matrix.getNumberOfIdentifiers(); ++indentifier_index ) {
208             final STATE_TYPE state = matrix.getState( indentifier_index, character_index );
209             if ( state == null ) {
210                 throw new IllegalArgumentException( "value at [" + indentifier_index + ", " + character_index
211                                                     + "] is null" );
212             }
213             final SortedSet<STATE_TYPE> l = new TreeSet<STATE_TYPE>();
214             l.add( state );
215             states.put( p.getNode( matrix.getIdentifier( indentifier_index ) ), l );
216         }
217         return states;
218     }
219
220     private Map<PhylogenyNode, STATE_TYPE> getStatesForCharacterForTraceback( final Phylogeny p,
221                                                                               final CharacterStateMatrix<STATE_TYPE> matrix,
222                                                                               final int character_index ) {
223         final Map<PhylogenyNode, STATE_TYPE> states = new HashMap<PhylogenyNode, STATE_TYPE>( matrix.getNumberOfIdentifiers() );
224         for( int indentifier_index = 0; indentifier_index < matrix.getNumberOfIdentifiers(); ++indentifier_index ) {
225             final STATE_TYPE state = matrix.getState( indentifier_index, character_index );
226             if ( state == null ) {
227                 throw new IllegalArgumentException( "value at [" + indentifier_index + ", " + character_index
228                                                     + "] is null" );
229             }
230             states.put( p.getNode( matrix.getIdentifier( indentifier_index ) ), state );
231         }
232         return states;
233     }
234
235     public int getTotalGains() {
236         return _total_gains;
237     }
238
239     public int getTotalLosses() {
240         return _total_losses;
241     }
242
243     public int getTotalUnchanged() {
244         return _total_unchanged;
245     }
246
247     private SortedSet<STATE_TYPE> getUnionOfStatesOfChildNodes( final Map<PhylogenyNode, SortedSet<STATE_TYPE>> states,
248                                                                 final PhylogenyNode node ) throws AssertionError {
249         final SortedSet<STATE_TYPE> states_in_child_nodes = new TreeSet<STATE_TYPE>();
250         for( int i = 0; i < node.getNumberOfDescendants(); ++i ) {
251             final PhylogenyNode node_child = node.getChildNode( i );
252             if ( !states.containsKey( node_child ) ) {
253                 throw new AssertionError( "this should not have happened: node [" + node_child.getName()
254                                           + "] not found in node state map" );
255             }
256             states_in_child_nodes.addAll( states.get( node_child ) );
257         }
258         return states_in_child_nodes;
259     }
260
261     private void increaseCost() {
262         ++_cost;
263     }
264
265     private void init() {
266         setReturnInternalStates( RETURN_INTERNAL_STATES_DEFAULT );
267         setReturnGainLossMatrix( RETURN_GAIN_LOSS_MATRIX_DEFAULT );
268         setRandomize( RANDOMIZE_DEFAULT );
269         setUseLast( USE_LAST_DEFAULT );
270         _random_number_seed = RANDOM_NUMBER_SEED_DEFAULT;
271         reset();
272     }
273
274     private void initializeGainLossMatrix( final Phylogeny p,
275                                            final CharacterStateMatrix<STATE_TYPE> external_node_states_matrix ) {
276         final List<PhylogenyNode> nodes = new ArrayList<PhylogenyNode>();
277         for( final PhylogenyNodeIterator postorder = p.iteratorPostorder(); postorder.hasNext(); ) {
278             nodes.add( postorder.next() );
279         }
280         setGainLossMatrix( new BasicCharacterStateMatrix<CharacterStateMatrix.GainLossStates>( nodes.size(),
281                 external_node_states_matrix
282                 .getNumberOfCharacters() ) );
283         int identifier_index = 0;
284         for( final PhylogenyNode node : nodes ) {
285             getGainLossMatrix().setIdentifier( identifier_index++,
286                                                ForesterUtil.isEmpty( node.getName() ) ? node.getId() + "" : node
287                                                        .getName() );
288         }
289         for( int character_index = 0; character_index < external_node_states_matrix.getNumberOfCharacters(); ++character_index ) {
290             getGainLossMatrix().setCharacter( character_index,
291                                               external_node_states_matrix.getCharacter( character_index ) );
292         }
293     }
294
295     private void initializeInternalStates( final Phylogeny p,
296                                            final CharacterStateMatrix<STATE_TYPE> external_node_states_matrix ) {
297         final List<PhylogenyNode> internal_nodes = new ArrayList<PhylogenyNode>();
298         for( final PhylogenyNodeIterator postorder = p.iteratorPostorder(); postorder.hasNext(); ) {
299             final PhylogenyNode node = postorder.next();
300             if ( node.isInternal() ) {
301                 internal_nodes.add( node );
302             }
303         }
304         setInternalStatesMatrixPriorToTraceback( new BasicCharacterStateMatrix<List<STATE_TYPE>>( internal_nodes.size(),
305                 external_node_states_matrix
306                 .getNumberOfCharacters() ) );
307         setInternalStatesMatrixTraceback( new BasicCharacterStateMatrix<STATE_TYPE>( internal_nodes.size(),
308                 external_node_states_matrix
309                 .getNumberOfCharacters() ) );
310         int identifier_index = 0;
311         for( final PhylogenyNode node : internal_nodes ) {
312             getInternalStatesMatrix().setIdentifier( identifier_index,
313                                                      ForesterUtil.isEmpty( node.getName() ) ? node.getId() + "" : node
314                                                              .getName() );
315             getInternalStatesMatrixPriorToTraceback().setIdentifier( identifier_index,
316                                                                      ForesterUtil.isEmpty( node.getName() ) ? node
317                                                                              .getId() + "" : node.getName() );
318             ++identifier_index;
319         }
320         for( int character_index = 0; character_index < external_node_states_matrix.getNumberOfCharacters(); ++character_index ) {
321             getInternalStatesMatrix().setCharacter( character_index,
322                                                     external_node_states_matrix.getCharacter( character_index ) );
323             getInternalStatesMatrixPriorToTraceback().setCharacter( character_index,
324                                                                     external_node_states_matrix
325                                                                     .getCharacter( character_index ) );
326         }
327     }
328
329     private boolean isRandomize() {
330         return _randomize;
331     }
332
333     private boolean isReturnGainLossMatrix() {
334         return _return_gain_loss;
335     }
336
337     private boolean isReturnInternalStates() {
338         return _return_internal_states;
339     }
340
341     private boolean isUseLast() {
342         return _use_last;
343     }
344
345     private void postOrderTraversal( final Phylogeny p, final Map<PhylogenyNode, SortedSet<STATE_TYPE>> states )
346             throws AssertionError {
347         for( final PhylogenyNodeIterator postorder = p.iteratorPostorder(); postorder.hasNext(); ) {
348             final PhylogenyNode node = postorder.next();
349             if ( !node.isExternal() ) {
350                 SortedSet<STATE_TYPE> states_in_children = getIntersectionOfStatesOfChildNodes( states, node );
351                 if ( states_in_children.isEmpty() ) {
352                     states_in_children = getUnionOfStatesOfChildNodes( states, node );
353                 }
354                 states.put( node, states_in_children );
355             }
356         }
357     }
358
359     private void preOrderTraversal( final Phylogeny p,
360                                     final Map<PhylogenyNode, SortedSet<STATE_TYPE>> states,
361                                     final Map<PhylogenyNode, STATE_TYPE> traceback_states,
362                                     final int character_state_column ) throws AssertionError {
363         for( final PhylogenyNodeIterator preorder = p.iteratorPreorder(); preorder.hasNext(); ) {
364             final PhylogenyNode current_node = preorder.next();
365             final SortedSet<STATE_TYPE> current_node_states = states.get( current_node );
366             STATE_TYPE parent_state = null;
367             if ( current_node.isRoot() ) {
368                 int i = 0;
369                 i = determineIndex( current_node_states, i );
370                 traceback_states.put( current_node, getStateAt( i, current_node_states ) );
371             }
372             else {
373                 parent_state = traceback_states.get( current_node.getParent() );
374                 if ( current_node_states.contains( parent_state ) ) {
375                     traceback_states.put( current_node, parent_state );
376                 }
377                 else {
378                     increaseCost();
379                     int i = 0;
380                     i = determineIndex( current_node_states, i );
381                     traceback_states.put( current_node, getStateAt( i, current_node_states ) );
382                 }
383             }
384             if ( isReturnInternalStates() ) {
385                 if ( !current_node.isExternal() ) {
386                     setInternalNodeStatePriorToTraceback( states, character_state_column, current_node );
387                     setInternalNodeState( traceback_states, character_state_column, current_node );
388                 }
389             }
390             if ( isReturnGainLossMatrix() && !current_node.isRoot() ) {
391                 if ( !( parent_state instanceof BinaryStates ) ) {
392                     throw new RuntimeException( "attempt to create gain loss matrix for not binary states" );
393                 }
394                 final BinaryStates parent_binary_state = ( BinaryStates ) parent_state;
395                 final BinaryStates current_binary_state = ( BinaryStates ) traceback_states.get( current_node );
396                 if ( ( parent_binary_state == PRESENT ) && ( current_binary_state == ABSENT ) ) {
397                     ++_total_losses;
398                     setGainLossState( character_state_column, current_node, LOSS );
399                 }
400                 else if ( ( ( parent_binary_state == ABSENT ) || ( parent_binary_state == null ) )
401                         && ( current_binary_state == PRESENT ) ) {
402                     ++_total_gains;
403                     setGainLossState( character_state_column, current_node, GAIN );
404                 }
405                 else {
406                     ++_total_unchanged;
407                     if ( current_binary_state == PRESENT ) {
408                         setGainLossState( character_state_column, current_node, UNCHANGED_PRESENT );
409                     }
410                     else if ( current_binary_state == ABSENT ) {
411                         setGainLossState( character_state_column, current_node, UNCHANGED_ABSENT );
412                     }
413                 }
414             }
415             else if ( isReturnGainLossMatrix() && current_node.isRoot() ) {
416                 final BinaryStates current_binary_state = ( BinaryStates ) traceback_states.get( current_node );
417                 ++_total_unchanged; //new
418                 if ( current_binary_state == PRESENT ) {//new
419                     setGainLossState( character_state_column, current_node, UNCHANGED_PRESENT );//new
420                 }//new
421                 else if ( current_binary_state == ABSENT ) {//new
422                     setGainLossState( character_state_column, current_node, UNCHANGED_ABSENT );//new
423                 }//new
424                 // setGainLossState( character_state_column, current_node, UNKNOWN_GAIN_LOSS );
425             }
426         }
427     }
428
429     private void reset() {
430         setCost( 0 );
431         setTotalLosses( 0 );
432         setTotalGains( 0 );
433         setTotalUnchanged( 0 );
434         setRandomGenerator( new Random( getRandomNumberSeed() ) );
435     }
436
437     private void setCost( final int cost ) {
438         _cost = cost;
439     }
440
441     private void setGainLossMatrix( final CharacterStateMatrix<GainLossStates> gain_loss_matrix ) {
442         _gain_loss_matrix = gain_loss_matrix;
443     }
444
445     private void setGainLossState( final int character_state_column,
446                                    final PhylogenyNode node,
447                                    final GainLossStates state ) {
448         getGainLossMatrix().setState( ForesterUtil.isEmpty( node.getName() ) ? node.getId() + "" : node.getName(),
449                 character_state_column,
450                 state );
451     }
452
453     private void setInternalNodeState( final Map<PhylogenyNode, STATE_TYPE> states,
454                                        final int character_state_column,
455                                        final PhylogenyNode node ) {
456         getInternalStatesMatrix()
457         .setState( ForesterUtil.isEmpty( node.getName() ) ? node.getId() + "" : node.getName(),
458                 character_state_column,
459                 states.get( node ) );
460     }
461
462     private void setInternalNodeStatePriorToTraceback( final Map<PhylogenyNode, SortedSet<STATE_TYPE>> states,
463                                                        final int character_state_column,
464                                                        final PhylogenyNode node ) {
465         getInternalStatesMatrixPriorToTraceback().setState( ForesterUtil.isEmpty( node.getName() ) ? node.getId() + ""
466                 : node.getName(),
467                 character_state_column,
468                 toListSorted( states.get( node ) ) );
469     }
470
471     private void setInternalStatesMatrixPriorToTraceback( final CharacterStateMatrix<List<STATE_TYPE>> internal_states_matrix_prior_to_traceback ) {
472         _internal_states_matrix_prior_to_traceback = internal_states_matrix_prior_to_traceback;
473     }
474
475     private void setInternalStatesMatrixTraceback( final CharacterStateMatrix<STATE_TYPE> internal_states_matrix_after_traceback ) {
476         _internal_states_matrix_after_traceback = internal_states_matrix_after_traceback;
477     }
478
479     private void setRandomGenerator( final Random random_generator ) {
480         _random_generator = random_generator;
481     }
482
483     public void setRandomize( final boolean randomize ) {
484         if ( randomize && isUseLast() ) {
485             throw new IllegalArgumentException( "attempt to allways use last state (ordered) if more than one choices and randomization at the same time" );
486         }
487         _randomize = randomize;
488     }
489
490     public void setRandomNumberSeed( final long random_number_seed ) {
491         if ( !isRandomize() ) {
492             throw new IllegalArgumentException( "attempt to set random number generator seed without randomization enabled" );
493         }
494         _random_number_seed = random_number_seed;
495     }
496
497     public void setReturnGainLossMatrix( final boolean return_gain_loss ) {
498         _return_gain_loss = return_gain_loss;
499     }
500
501     public void setReturnInternalStates( final boolean return_internal_states ) {
502         _return_internal_states = return_internal_states;
503     }
504
505     private void setTotalGains( final int total_gains ) {
506         _total_gains = total_gains;
507     }
508
509     private void setTotalLosses( final int total_losses ) {
510         _total_losses = total_losses;
511     }
512
513     private void setTotalUnchanged( final int total_unchanged ) {
514         _total_unchanged = total_unchanged;
515     }
516
517     /**
518      * This sets whether to use the first or last state in the sorted
519      * states at the undecided internal nodes.
520      * For randomized choices set randomize to true (and this to false).
521      *
522      * Note. It might be advisable to set this to false
523      * for BinaryStates if absence at the root is preferred
524      * (given the enum BinaryStates sorts in the following order:
525      * ABSENT, UNKNOWN, PRESENT).
526      *
527      *
528      * @param use_last
529      */
530     public void setUseLast( final boolean use_last ) {
531         if ( use_last && isRandomize() ) {
532             throw new IllegalArgumentException( "attempt to allways use last state (ordered) if more than one choices and randomization at the same time" );
533         }
534         _use_last = use_last;
535     }
536
537     private List<STATE_TYPE> toListSorted( final SortedSet<STATE_TYPE> states ) {
538         final List<STATE_TYPE> l = new ArrayList<STATE_TYPE>( states.size() );
539         for( final STATE_TYPE state : states ) {
540             l.add( state );
541         }
542         return l;
543     }
544 }