[1573] | 1 | // UPGMATree.java |
---|
| 2 | // |
---|
| 3 | // (c) 1999-2001 PAL Development Core Team |
---|
| 4 | // |
---|
| 5 | // This package may be distributed under the |
---|
| 6 | // terms of the Lesser GNU General Public License (LGPL) |
---|
| 7 | |
---|
| 8 | // Known bugs and limitations: |
---|
| 9 | // - computational complexity O(numSeqs^3) |
---|
| 10 | // (this could be brought down to O(numSeqs^2) |
---|
| 11 | // but this needs more clever programming ...) |
---|
| 12 | |
---|
| 13 | |
---|
| 14 | package de.ugoe.cs.autoquest.tasktrees.alignment.pal.tree; |
---|
| 15 | |
---|
| 16 | import java.util.ArrayList; |
---|
[1585] | 17 | import java.util.Iterator; |
---|
| 18 | import java.util.logging.Level; |
---|
[1573] | 19 | |
---|
| 20 | import de.ugoe.cs.autoquest.tasktrees.alignment.algorithms.NumberSequence; |
---|
[1585] | 21 | import de.ugoe.cs.autoquest.tasktrees.alignment.algorithms.SmithWatermanRepeated; |
---|
[1584] | 22 | import de.ugoe.cs.autoquest.tasktrees.alignment.matrix.BinaryAlignmentStorage; |
---|
[1585] | 23 | import de.ugoe.cs.autoquest.tasktrees.alignment.matrix.ObjectDistanceSubstitionMatrix; |
---|
[1573] | 24 | import de.ugoe.cs.autoquest.tasktrees.alignment.matrix.UPGMAMatrix; |
---|
| 25 | import de.ugoe.cs.autoquest.tasktrees.alignment.pal.misc.Identifier; |
---|
[1585] | 26 | import de.ugoe.cs.util.console.Console; |
---|
[1573] | 27 | |
---|
| 28 | |
---|
| 29 | /** |
---|
| 30 | * constructs a UPGMA tree from pairwise distances |
---|
| 31 | * |
---|
| 32 | * @version $Id: UPGMATree.java,v 1.9 2001/07/13 14:39:13 korbinian Exp $ |
---|
| 33 | * |
---|
| 34 | * @author Korbinian Strimmer |
---|
| 35 | * @author Alexei Drummond |
---|
| 36 | */ |
---|
[1579] | 37 | public class UPGMAAligningTree extends SimpleTree |
---|
[1573] | 38 | { |
---|
| 39 | // |
---|
| 40 | // Public stuff |
---|
| 41 | // |
---|
| 42 | |
---|
| 43 | /** |
---|
| 44 | * constructor UPGMA tree |
---|
| 45 | * @param numberseqs |
---|
| 46 | * |
---|
| 47 | * @param m distance matrix |
---|
| 48 | */ |
---|
[1585] | 49 | public UPGMAAligningTree(ArrayList<NumberSequence> numberseqs, BinaryAlignmentStorage alignments, ObjectDistanceSubstitionMatrix submat) |
---|
[1573] | 50 | { |
---|
[1584] | 51 | if (alignments.getDistanceMatrix().size() < 2) |
---|
[1573] | 52 | { |
---|
| 53 | new IllegalArgumentException("LESS THAN 2 TAXA IN DISTANCE MATRIX"); |
---|
| 54 | } |
---|
| 55 | |
---|
[1577] | 56 | this.numberseqs = numberseqs; |
---|
[1584] | 57 | this.alignments = alignments; |
---|
[1585] | 58 | this.submat = submat; |
---|
[1584] | 59 | init(alignments.getDistanceMatrix()); |
---|
[1573] | 60 | |
---|
| 61 | while (true) |
---|
| 62 | { |
---|
| 63 | findNextPair(); |
---|
| 64 | newBranchLengths(); |
---|
| 65 | |
---|
| 66 | if (numClusters == 2) |
---|
| 67 | { |
---|
| 68 | break; |
---|
| 69 | } |
---|
| 70 | |
---|
| 71 | newCluster(); |
---|
| 72 | } |
---|
| 73 | |
---|
| 74 | finish(); |
---|
| 75 | createNodeList(); |
---|
| 76 | } |
---|
| 77 | |
---|
| 78 | |
---|
| 79 | // |
---|
| 80 | // Private stuff |
---|
| 81 | // |
---|
| 82 | private ArrayList<NumberSequence> numberseqs; |
---|
[1584] | 83 | private BinaryAlignmentStorage alignments; |
---|
[1585] | 84 | private ObjectDistanceSubstitionMatrix submat; |
---|
[1573] | 85 | private int numClusters; |
---|
| 86 | private int besti, abi; |
---|
| 87 | private int bestj, abj; |
---|
| 88 | private int[] alias; |
---|
| 89 | private double[][] distance; |
---|
| 90 | |
---|
| 91 | private double[] height; |
---|
| 92 | private int[] oc; |
---|
| 93 | |
---|
| 94 | private double getDist(int a, int b) |
---|
| 95 | { |
---|
| 96 | return distance[alias[a]][alias[b]]; |
---|
| 97 | } |
---|
| 98 | |
---|
| 99 | private void init(UPGMAMatrix m) |
---|
| 100 | { |
---|
| 101 | numClusters = m.size(); |
---|
| 102 | |
---|
| 103 | distance = new double[numClusters][numClusters]; |
---|
| 104 | for (int i = 0; i < numClusters; i++) |
---|
| 105 | { |
---|
| 106 | for (int j = 0; j < numClusters; j++) |
---|
| 107 | { |
---|
| 108 | distance[i][j] = m.get(i,j); |
---|
| 109 | } |
---|
| 110 | } |
---|
| 111 | |
---|
| 112 | for (int i = 0; i < numClusters; i++) |
---|
| 113 | { |
---|
| 114 | Node tmp = NodeFactory.createNode(); |
---|
| 115 | tmp.setIdentifier(new Identifier(Integer.toString(i))); |
---|
[1585] | 116 | tmp.setNumber(i); |
---|
[1577] | 117 | tmp.addSequence(numberseqs.get(i)); |
---|
[1573] | 118 | getRoot().addChild(tmp); |
---|
| 119 | } |
---|
| 120 | |
---|
| 121 | alias = new int[numClusters]; |
---|
| 122 | for (int i = 0; i < numClusters; i++) |
---|
| 123 | { |
---|
| 124 | alias[i] = i; |
---|
| 125 | } |
---|
| 126 | |
---|
| 127 | height = new double[numClusters]; |
---|
| 128 | oc = new int[numClusters]; |
---|
| 129 | for (int i = 0; i < numClusters; i++) |
---|
| 130 | { |
---|
| 131 | height[i] = 0.0; |
---|
| 132 | oc[i] = 1; |
---|
| 133 | } |
---|
| 134 | } |
---|
| 135 | |
---|
| 136 | private void finish() |
---|
| 137 | { |
---|
| 138 | distance = null; |
---|
| 139 | } |
---|
| 140 | |
---|
| 141 | private void findNextPair() |
---|
| 142 | { |
---|
| 143 | besti = 0; |
---|
| 144 | bestj = 1; |
---|
| 145 | double dmin = getDist(0, 1); |
---|
| 146 | for (int i = 0; i < numClusters-1; i++) |
---|
| 147 | { |
---|
| 148 | for (int j = i+1; j < numClusters; j++) |
---|
| 149 | { |
---|
| 150 | if (getDist(i, j) < dmin) |
---|
| 151 | { |
---|
| 152 | dmin = getDist(i, j); |
---|
| 153 | besti = i; |
---|
| 154 | bestj = j; |
---|
| 155 | } |
---|
| 156 | } |
---|
| 157 | } |
---|
| 158 | abi = alias[besti]; |
---|
| 159 | abj = alias[bestj]; |
---|
[1577] | 160 | //System.out.println("Found best pair: " + abi + "/" +abj + " - "+ besti+ "/"+bestj +" with distance " + dmin); |
---|
[1573] | 161 | |
---|
| 162 | } |
---|
| 163 | |
---|
| 164 | private void newBranchLengths() |
---|
| 165 | { |
---|
| 166 | double dij = getDist(besti, bestj); |
---|
| 167 | |
---|
| 168 | getRoot().getChild(besti).setBranchLength(dij/2.0-height[abi]); |
---|
| 169 | getRoot().getChild(bestj).setBranchLength(dij/2.0-height[abj]); |
---|
| 170 | } |
---|
| 171 | |
---|
| 172 | private void newCluster() |
---|
| 173 | { |
---|
| 174 | // Update distances |
---|
| 175 | for (int k = 0; k < numClusters; k++) |
---|
| 176 | { |
---|
| 177 | if (k != besti && k != bestj) |
---|
| 178 | { |
---|
| 179 | int ak = alias[k]; |
---|
| 180 | double updated = updatedDistance(besti,bestj,k); |
---|
| 181 | distance[ak][abi] = distance[abi][ak] = updated; |
---|
| 182 | } |
---|
| 183 | } |
---|
| 184 | distance[abi][abi] = 0.0; |
---|
| 185 | |
---|
| 186 | // Update UPGMA variables |
---|
| 187 | height[abi] = getDist(besti, bestj)/2.0; |
---|
| 188 | oc[abi] += oc[abj]; |
---|
| 189 | |
---|
| 190 | // Index besti now represent the new cluster |
---|
[1585] | 191 | Node newNode = getRoot().joinChildren(besti, bestj); |
---|
[1573] | 192 | |
---|
[1585] | 193 | if(newNode instanceof FengDoolittleNode) { |
---|
| 194 | newNode.setSequences(alignSequences(newNode)); |
---|
| 195 | } |
---|
| 196 | |
---|
[1573] | 197 | // Update alias |
---|
| 198 | for (int i = bestj; i < numClusters-1; i++) |
---|
| 199 | { |
---|
| 200 | alias[i] = alias[i+1]; |
---|
| 201 | } |
---|
| 202 | |
---|
| 203 | numClusters--; |
---|
| 204 | } |
---|
[1585] | 205 | |
---|
| 206 | |
---|
| 207 | public ArrayList<NumberSequence> alignSequences(Node parent) { |
---|
| 208 | ArrayList<NumberSequence> alignment = new ArrayList<NumberSequence>(); |
---|
| 209 | if(parent.getChildCount()<3) { |
---|
| 210 | |
---|
| 211 | Node node1 = parent.getChild(0); |
---|
| 212 | Node node2 = parent.getChild(1); |
---|
| 213 | |
---|
| 214 | int seqCount1 = node1.getSequences().size(); |
---|
| 215 | int seqCount2 = node2.getSequences().size(); |
---|
| 216 | Console.traceln(Level.INFO,"Merging node " + node1.getIdentifier() + " with " + node2.getIdentifier()); |
---|
| 217 | //Console.println("SeqCount1: " + seqCount1 + " seqCount2 " + seqCount2); |
---|
| 218 | //Align 2 sequences |
---|
| 219 | if(seqCount1 == 1 && seqCount2 == 1) { |
---|
| 220 | alignment = (alignments.get(node1.getNumber(), node2.getNumber())).getAlignment(); |
---|
| 221 | } |
---|
| 222 | //Align a sequence to a group |
---|
| 223 | else if( seqCount1 > 1 && seqCount2 == 1) { |
---|
| 224 | alignment.addAll(node1.getSequences()); |
---|
| 225 | |
---|
| 226 | BinaryAlignmentStorage tempStorage = new BinaryAlignmentStorage(seqCount1,seqCount2); |
---|
| 227 | double maxScore = 0.0; |
---|
| 228 | int maxIndex = 0; |
---|
| 229 | for(int i=0;i<seqCount1;i++) { |
---|
| 230 | tempStorage.set(i, 1, new SmithWatermanRepeated(node1.getSequence(i).getSequence(), node2.getSequence(0).getSequence() , submat, 5)); |
---|
| 231 | if(maxScore < tempStorage.get(i, 1).getAlignmentScore()) { |
---|
| 232 | maxScore = tempStorage.get(i, 1).getAlignmentScore(); |
---|
| 233 | maxIndex = i; |
---|
| 234 | } |
---|
| 235 | } |
---|
| 236 | alignment.add(tempStorage.get(maxIndex, 1).getAlignment().get(1)); |
---|
| 237 | } |
---|
| 238 | //Align a sequence to a group |
---|
| 239 | else if(seqCount1 == 1 && seqCount2 > 1) { |
---|
| 240 | alignment.addAll(node2.getSequences()); |
---|
| 241 | BinaryAlignmentStorage tempStorage = new BinaryAlignmentStorage(seqCount1,seqCount2); |
---|
| 242 | double maxScore = 0.0; |
---|
| 243 | int maxIndex = 0; |
---|
| 244 | for(int i=0;i<seqCount2;i++) { |
---|
| 245 | tempStorage.set(1, i, new SmithWatermanRepeated(node2.getSequence(i).getSequence(), node1.getSequence(0).getSequence() , submat, 5)); |
---|
| 246 | if(maxScore < tempStorage.get(1, i).getAlignmentScore()) { |
---|
| 247 | maxScore = tempStorage.get(1, i).getAlignmentScore(); |
---|
| 248 | maxIndex = i; |
---|
| 249 | } |
---|
| 250 | } |
---|
| 251 | alignment.add(tempStorage.get(1,maxIndex).getAlignment().get(1)); |
---|
| 252 | } |
---|
| 253 | //Align 2 groups |
---|
| 254 | else if((seqCount1 > 1) && (seqCount2 > 1)){ |
---|
| 255 | BinaryAlignmentStorage tempStorage1 = new BinaryAlignmentStorage(seqCount2,1); |
---|
| 256 | BinaryAlignmentStorage tempStorage2 = new BinaryAlignmentStorage(seqCount1,1); |
---|
| 257 | double maxScore1 = 0.0; |
---|
| 258 | double maxScore2 = 0.0; |
---|
| 259 | int maxIndex1 = 0; |
---|
| 260 | int maxIndex2 = 0; |
---|
| 261 | for(int i=0;i<seqCount1;i++) { |
---|
| 262 | for(int j=0;j<seqCount2;j++) { |
---|
| 263 | tempStorage1.set(j, 0, new SmithWatermanRepeated(node1.getSequence(i).getSequence(), node2.getSequence(j).getSequence() , submat, 5)); |
---|
| 264 | if(maxScore1 < tempStorage1.get(j, 0).getAlignmentScore()) { |
---|
| 265 | maxScore1 = tempStorage1.get(j, 0).getAlignmentScore(); |
---|
| 266 | maxIndex1 = j; |
---|
| 267 | } |
---|
| 268 | } |
---|
| 269 | alignment.add(tempStorage1.get(maxIndex1,0).getAlignment().get(0)); |
---|
| 270 | } |
---|
| 271 | for(int i=0; i<seqCount2;i++) { |
---|
| 272 | for (int j=0;j<seqCount1;j++) { |
---|
| 273 | tempStorage2.set(j, 0, new SmithWatermanRepeated(node2.getSequence(i).getSequence(),node1.getSequence(j).getSequence(),submat,5)); |
---|
| 274 | if(maxScore2 < tempStorage2.get(j, 0).getAlignmentScore()) { |
---|
| 275 | maxScore2 = tempStorage2.get(j, 0).getAlignmentScore(); |
---|
| 276 | maxIndex2 = j; |
---|
| 277 | } |
---|
| 278 | } |
---|
| 279 | alignment.add(tempStorage2.get(maxIndex2,0).getAlignment().get(0)); |
---|
| 280 | } |
---|
| 281 | |
---|
| 282 | } |
---|
| 283 | else { |
---|
| 284 | Console.traceln(Level.WARNING,"No sequences to align while merging " + node1.getIdentifier() + " with " + node2.getIdentifier()); |
---|
| 285 | } |
---|
| 286 | } |
---|
| 287 | else { |
---|
| 288 | Console.traceln(Level.WARNING,"More than 2 children! This should never happen, it's a binary tree."); |
---|
| 289 | } |
---|
| 290 | return alignment; |
---|
| 291 | } |
---|
[1573] | 292 | |
---|
| 293 | |
---|
| 294 | /** |
---|
| 295 | * compute updated distance between the new cluster (i,j) |
---|
| 296 | * to any other cluster k |
---|
| 297 | */ |
---|
| 298 | private double updatedDistance(int i, int j, int k) |
---|
| 299 | { |
---|
| 300 | int ai = alias[i]; |
---|
| 301 | int aj = alias[j]; |
---|
| 302 | |
---|
| 303 | double ocsum = (double) (oc[ai]+oc[aj]); |
---|
| 304 | double idist = getDist(k,i); |
---|
| 305 | double jdist = getDist(k,j); |
---|
| 306 | //TODO: Dirty hack to deal with infinity, insert proper solution here |
---|
| 307 | if(Double.isInfinite(idist)) { |
---|
| 308 | idist = 100; |
---|
| 309 | } |
---|
| 310 | if(Double.isInfinite(jdist)) { |
---|
| 311 | jdist = 100; |
---|
| 312 | } |
---|
| 313 | |
---|
| 314 | return (oc[ai]/ocsum)*idist+ |
---|
| 315 | (oc[aj]/ocsum)*jdist; |
---|
| 316 | } |
---|
| 317 | } |
---|
[1585] | 318 | |
---|
| 319 | |
---|