[1587] | 1 | package de.ugoe.cs.autoquest.tasktrees.alignment.algorithms; |
---|
| 2 | |
---|
| 3 | import java.util.ArrayList; |
---|
| 4 | import java.util.logging.Level; |
---|
| 5 | |
---|
| 6 | import de.ugoe.cs.autoquest.tasktrees.alignment.matrix.SubstitutionMatrix; |
---|
| 7 | import de.ugoe.cs.autoquest.tasktrees.alignment.algorithms.Constants; |
---|
| 8 | import de.ugoe.cs.util.console.Console; |
---|
| 9 | |
---|
| 10 | public class NeedlemanWunsch implements AlignmentAlgorithm { |
---|
| 11 | |
---|
| 12 | /** |
---|
| 13 | * The first input |
---|
| 14 | */ |
---|
| 15 | private int[] input1; |
---|
| 16 | |
---|
| 17 | /** |
---|
| 18 | * The second input String |
---|
| 19 | */ |
---|
| 20 | private int[] input2; |
---|
| 21 | |
---|
| 22 | /** |
---|
| 23 | * The lengths of the input |
---|
| 24 | */ |
---|
| 25 | private int length1, length2; |
---|
| 26 | |
---|
| 27 | /** |
---|
| 28 | * The score matrix. The true scores should be divided by the normalization |
---|
| 29 | * factor. |
---|
| 30 | */ |
---|
| 31 | private MatrixEntry[][] matrix; |
---|
| 32 | |
---|
| 33 | private ArrayList<NumberSequence> alignment; |
---|
| 34 | |
---|
| 35 | /** |
---|
| 36 | * Substitution matrix to calculate scores |
---|
| 37 | */ |
---|
| 38 | private SubstitutionMatrix submat; |
---|
| 39 | |
---|
| 40 | public NeedlemanWunsch(int[] input1, int[] input2, SubstitutionMatrix submat, |
---|
| 41 | float threshold) { |
---|
| 42 | this.input1 = input1; |
---|
| 43 | this.input2 = input2; |
---|
| 44 | length1 = input1.length; |
---|
| 45 | length2 = input2.length; |
---|
| 46 | this.submat = submat; |
---|
| 47 | |
---|
| 48 | // System.out.println("Starting SmithWaterman algorithm with a " |
---|
| 49 | // + submat.getClass() + " Substitution Matrix: " + |
---|
| 50 | // submat.getClass().getCanonicalName()); |
---|
| 51 | |
---|
| 52 | matrix = new MatrixEntry[length1 + 1][length2 + 1]; |
---|
| 53 | alignment = new ArrayList<NumberSequence>(); |
---|
| 54 | |
---|
| 55 | for (int i = 0; i < length1+1; i++) { |
---|
| 56 | for (int j = 0; j < length2+1; j++) { |
---|
| 57 | matrix[i][j] = new MatrixEntry(); |
---|
| 58 | } |
---|
| 59 | } |
---|
| 60 | |
---|
| 61 | buildMatrix(); |
---|
| 62 | traceback(); |
---|
| 63 | } |
---|
| 64 | |
---|
| 65 | /** |
---|
| 66 | * Compute the similarity score of substitution The position of the first |
---|
| 67 | * character is 1. A position of 0 represents a gap. |
---|
| 68 | * |
---|
| 69 | * @param i |
---|
| 70 | * Position of the character in str1 |
---|
| 71 | * @param j |
---|
| 72 | * Position of the character in str2 |
---|
| 73 | * @return Cost of substitution of the character in str1 by the one in str2 |
---|
| 74 | */ |
---|
| 75 | private double similarity(int i, int j) { |
---|
| 76 | return submat.getScore(input1[i - 1], input2[j - 1]); |
---|
| 77 | } |
---|
| 78 | |
---|
| 79 | /** |
---|
| 80 | * Build the score matrix using dynamic programming. |
---|
| 81 | */ |
---|
| 82 | private void buildMatrix() { |
---|
| 83 | if (submat.getGapPenalty() >= 0) { |
---|
| 84 | throw new Error("Indel score must be negative"); |
---|
| 85 | } |
---|
| 86 | |
---|
| 87 | // it's a gap |
---|
| 88 | matrix[0][0].setScore(0); |
---|
| 89 | matrix[0][0].setPrevious(null); // starting point |
---|
| 90 | |
---|
| 91 | // the first column |
---|
| 92 | for (int j = 1; j <= length2; j++) { |
---|
| 93 | matrix[0][j].setScore(j*submat.getGapPenalty()); |
---|
| 94 | matrix[0][j].setPrevious(matrix[0][j - 1]); |
---|
| 95 | matrix[0][j].setYvalue(input2[j-1]); |
---|
| 96 | matrix[0][j].setXvalue(Constants.GAP_SYMBOL); |
---|
| 97 | } |
---|
| 98 | // the first row |
---|
| 99 | |
---|
| 100 | for (int j = 1; j <= length1; j++) { |
---|
| 101 | matrix[j][0].setScore(j*submat.getGapPenalty()); |
---|
| 102 | matrix[j][0].setPrevious(matrix[j - 1][0]); |
---|
| 103 | matrix[j][0].setXvalue(input1[j-1]); |
---|
| 104 | matrix[j][0].setYvalue(Constants.GAP_SYMBOL); |
---|
| 105 | } |
---|
| 106 | |
---|
| 107 | for (int i = 1; i <= length1; i++) { |
---|
| 108 | |
---|
| 109 | for (int j = 1; j <= length2; j++) { |
---|
| 110 | double diagScore = matrix[i - 1][j - 1].getScore() |
---|
| 111 | + similarity(i, j); |
---|
| 112 | double upScore = matrix[i][j - 1].getScore() |
---|
| 113 | + submat.getGapPenalty(); |
---|
| 114 | double leftScore = matrix[i - 1][j].getScore() |
---|
| 115 | + submat.getGapPenalty(); |
---|
| 116 | |
---|
| 117 | matrix[i][j].setScore(Math.max(diagScore, |
---|
| 118 | Math.max(upScore, Math.max(leftScore, 0)))); |
---|
| 119 | |
---|
| 120 | // find the directions that give the maximum scores. |
---|
| 121 | // TODO: Multiple directions are ignored, we choose the first |
---|
| 122 | // maximum score |
---|
| 123 | // True if we had a match |
---|
| 124 | if (diagScore == matrix[i][j].getScore()) { |
---|
| 125 | matrix[i][j].setPrevious(matrix[i - 1][j - 1]); |
---|
| 126 | matrix[i][j].setXvalue(input1[i - 1]); |
---|
| 127 | matrix[i][j].setYvalue(input2[j - 1]); |
---|
| 128 | } |
---|
| 129 | // true if we took an event from sequence x and not from y |
---|
| 130 | if (leftScore == matrix[i][j].getScore()) { |
---|
| 131 | matrix[i][j].setXvalue(input1[i - 1]); |
---|
| 132 | matrix[i][j].setYvalue(Constants.GAP_SYMBOL); |
---|
| 133 | matrix[i][j].setPrevious(matrix[i - 1][j]); |
---|
| 134 | } |
---|
| 135 | // true if we took an event from sequence y and not from x |
---|
| 136 | if (upScore == matrix[i][j].getScore()) { |
---|
| 137 | matrix[i][j].setXvalue(Constants.GAP_SYMBOL); |
---|
| 138 | matrix[i][j].setYvalue(input2[j - 1]); |
---|
| 139 | matrix[i][j].setPrevious(matrix[i][j - 1]); |
---|
| 140 | } |
---|
| 141 | } |
---|
| 142 | } |
---|
| 143 | } |
---|
| 144 | |
---|
| 145 | /** |
---|
| 146 | * Get the maximum value in the score matrix. |
---|
| 147 | */ |
---|
| 148 | public double getMaxScore() { |
---|
| 149 | double maxScore = 0; |
---|
| 150 | |
---|
| 151 | // skip the first row and column |
---|
| 152 | for (int i = 1; i <= length1; i++) { |
---|
| 153 | for (int j = 1; j <= length2; j++) { |
---|
| 154 | if (matrix[i][j].getScore() > maxScore) { |
---|
| 155 | maxScore = matrix[i][j].getScore(); |
---|
| 156 | } |
---|
| 157 | } |
---|
| 158 | } |
---|
| 159 | |
---|
| 160 | return maxScore; |
---|
| 161 | } |
---|
| 162 | |
---|
| 163 | /* |
---|
| 164 | * (non-Javadoc) |
---|
| 165 | * |
---|
| 166 | * @see |
---|
| 167 | * de.ugoe.cs.autoquest.tasktrees.alignment.algorithms.AlignmentAlgorithm |
---|
| 168 | * #getAlignmentScore() |
---|
| 169 | */ |
---|
| 170 | @Override |
---|
| 171 | public double getAlignmentScore() { |
---|
| 172 | return getMaxScore(); |
---|
| 173 | } |
---|
| 174 | |
---|
| 175 | public void traceback() { |
---|
| 176 | MatrixEntry tmp = matrix[length1][length2]; |
---|
| 177 | int aligned1[] = new int[length1 + length2 + 2]; |
---|
| 178 | int aligned2[] = new int[length1 + length2 + 2]; |
---|
| 179 | int count = 0; |
---|
| 180 | do { |
---|
| 181 | if (length1 + length2 + 2 == count) { |
---|
| 182 | Console.traceln(Level.WARNING, |
---|
| 183 | "Traceback longer than both sequences summed up!"); |
---|
| 184 | break; |
---|
| 185 | } |
---|
| 186 | aligned1[count] = tmp.getXvalue(); |
---|
| 187 | aligned2[count] = tmp.getYvalue(); |
---|
| 188 | |
---|
| 189 | tmp = tmp.getPrevious(); |
---|
| 190 | count++; |
---|
| 191 | |
---|
| 192 | } while (tmp != null); |
---|
| 193 | count--; |
---|
| 194 | // reverse order of the alignment |
---|
| 195 | int reversed1[] = new int[count]; |
---|
| 196 | int reversed2[] = new int[count]; |
---|
| 197 | |
---|
| 198 | for (int i = count; i > 0; i--) { |
---|
| 199 | reversed1[reversed1.length - i] = aligned1[i]; |
---|
| 200 | reversed2[reversed2.length - i] = aligned2[i]; |
---|
| 201 | } |
---|
| 202 | |
---|
| 203 | NumberSequence ns1 = new NumberSequence(reversed1.length); |
---|
| 204 | NumberSequence ns2 = new NumberSequence(reversed2.length); |
---|
| 205 | ns1.setSequence(reversed1); |
---|
| 206 | ns2.setSequence(reversed2); |
---|
| 207 | |
---|
| 208 | alignment.add(ns1); |
---|
| 209 | alignment.add(ns2); |
---|
| 210 | } |
---|
| 211 | |
---|
| 212 | public void printAlignment() { |
---|
| 213 | MatrixEntry tmp = matrix[length1][length2]; |
---|
| 214 | String aligned1 = ""; |
---|
| 215 | String aligned2 = ""; |
---|
| 216 | int count = 0; |
---|
| 217 | do { |
---|
| 218 | String append1 = ""; |
---|
| 219 | String append2 = ""; |
---|
| 220 | |
---|
| 221 | if (tmp.getXvalue() == Constants.GAP_SYMBOL) { |
---|
| 222 | append1 = " ___"; |
---|
| 223 | } else { |
---|
| 224 | append1 = String.format("%5d", tmp.getXvalue()); |
---|
| 225 | } |
---|
| 226 | |
---|
| 227 | if (tmp.getYvalue() == Constants.GAP_SYMBOL) { |
---|
| 228 | append2 = " ___"; |
---|
| 229 | } else { |
---|
| 230 | append2 = String.format("%5d", tmp.getYvalue()); |
---|
| 231 | } |
---|
| 232 | if (count != 0) { |
---|
| 233 | aligned1 = append1 + aligned1; |
---|
| 234 | aligned2 = append2 + aligned2; |
---|
| 235 | } |
---|
| 236 | |
---|
| 237 | tmp = tmp.getPrevious(); |
---|
| 238 | count++; |
---|
| 239 | |
---|
| 240 | } while (tmp != null); |
---|
| 241 | System.out.println(aligned1); |
---|
| 242 | System.out.println(aligned2); |
---|
| 243 | } |
---|
| 244 | |
---|
| 245 | /** |
---|
| 246 | * print the dynmaic programming matrix |
---|
| 247 | */ |
---|
| 248 | public void printDPMatrix() { |
---|
| 249 | System.out.print(" "); |
---|
| 250 | for (int i = 1; i <= length1; i++) |
---|
| 251 | System.out.format("%5d", input1[i - 1]); |
---|
| 252 | System.out.println(); |
---|
| 253 | for (int j = 0; j <= length2; j++) { |
---|
| 254 | if (j > 0) |
---|
| 255 | System.out.format("%5d ", input2[j - 1]); |
---|
| 256 | else { |
---|
| 257 | System.out.print(" "); |
---|
| 258 | } |
---|
| 259 | for (int i = 0; i <= length1; i++) { |
---|
| 260 | System.out.format("%4.1f ", matrix[i][j].getScore()); |
---|
| 261 | } |
---|
| 262 | System.out.println(); |
---|
| 263 | } |
---|
| 264 | } |
---|
| 265 | |
---|
| 266 | /* |
---|
| 267 | * (non-Javadoc) |
---|
| 268 | * |
---|
| 269 | * @see |
---|
| 270 | * de.ugoe.cs.autoquest.tasktrees.alignment.algorithms.AlignmentAlgorithm |
---|
| 271 | * #getAlignment() |
---|
| 272 | */ |
---|
| 273 | @Override |
---|
| 274 | public ArrayList<NumberSequence> getAlignment() { |
---|
| 275 | return alignment; |
---|
| 276 | } |
---|
| 277 | |
---|
| 278 | public void setAlignment(ArrayList<NumberSequence> alignment) { |
---|
| 279 | this.alignment = alignment; |
---|
| 280 | } |
---|
| 281 | |
---|
| 282 | |
---|
| 283 | |
---|
| 284 | } |
---|