1 | package de.ugoe.cs.autoquest.tasktrees.alignment.algorithms; |
---|
2 | |
---|
3 | import java.util.ArrayList; |
---|
4 | import java.util.Iterator; |
---|
5 | import java.util.LinkedList; |
---|
6 | import java.util.List; |
---|
7 | |
---|
8 | import de.ugoe.cs.autoquest.tasktrees.alignment.matrix.SubstitutionMatrix; |
---|
9 | |
---|
10 | public class SmithWatermanRepeated { |
---|
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 | |
---|
34 | private List<NumberSequence> alignment; |
---|
35 | |
---|
36 | private float scoreThreshold; |
---|
37 | |
---|
38 | /** |
---|
39 | * Substitution matrix to calculate scores |
---|
40 | */ |
---|
41 | private SubstitutionMatrix submat; |
---|
42 | |
---|
43 | public SmithWatermanRepeated(int[] input1, int[] input2, SubstitutionMatrix submat,float threshold) { |
---|
44 | this.input1 = input1; |
---|
45 | this.input2 = input2; |
---|
46 | length1 = input1.length; |
---|
47 | length2 = input2.length; |
---|
48 | this.submat = submat; |
---|
49 | |
---|
50 | //System.out.println("Starting SmithWaterman algorithm with a " |
---|
51 | // + submat.getClass() + " Substitution Matrix: " + submat.getClass().getCanonicalName()); |
---|
52 | this.scoreThreshold = threshold; |
---|
53 | |
---|
54 | matrix = new MatrixEntry[length1+2][length2+1]; |
---|
55 | alignment = new ArrayList<NumberSequence>(); |
---|
56 | |
---|
57 | for (int i = 0; i <= length1+1; i++) { |
---|
58 | for(int j = 0; j< length2; j++) { |
---|
59 | matrix[i][j] = new MatrixEntry(); |
---|
60 | } |
---|
61 | } |
---|
62 | |
---|
63 | |
---|
64 | buildMatrix(); |
---|
65 | traceback(); |
---|
66 | } |
---|
67 | |
---|
68 | /** |
---|
69 | * Compute the similarity score of substitution The position of the first |
---|
70 | * character is 1. A position of 0 represents a gap. |
---|
71 | * |
---|
72 | * @param i |
---|
73 | * Position of the character in str1 |
---|
74 | * @param j |
---|
75 | * Position of the character in str2 |
---|
76 | * @return Cost of substitution of the character in str1 by the one in str2 |
---|
77 | */ |
---|
78 | private double similarity(int i, int j) { |
---|
79 | return submat.getDistance(input1[i - 1], input2[j - 1]); |
---|
80 | } |
---|
81 | |
---|
82 | /** |
---|
83 | * Build the score matrix using dynamic programming. |
---|
84 | */ |
---|
85 | private void buildMatrix() { |
---|
86 | if (submat.getGapPenalty() >= 0) { |
---|
87 | throw new Error("Indel score must be negative"); |
---|
88 | } |
---|
89 | |
---|
90 | // it's a gap |
---|
91 | matrix[0][0].setScore(0); |
---|
92 | matrix[0][0].setPrevious(null); // starting point |
---|
93 | |
---|
94 | // the first column |
---|
95 | for (int j = 1; j < length2; j++) { |
---|
96 | matrix[0][j].setScore(0); |
---|
97 | matrix[0][j].setPrevious(matrix[0][j-1]); |
---|
98 | |
---|
99 | } |
---|
100 | |
---|
101 | |
---|
102 | |
---|
103 | for (int i = 1; i <= length1 + 1; i++) { |
---|
104 | |
---|
105 | // Formula for first row: |
---|
106 | // F(i,0) = max { F(i-1,0), F(i-1,j)-T j=1,...,m |
---|
107 | |
---|
108 | double firstRowLeftScore = matrix[i-1][0].getScore(); |
---|
109 | //for sequences of length 1 |
---|
110 | double tempMax; |
---|
111 | int maxRowIndex; |
---|
112 | if(length2 == 1) { |
---|
113 | tempMax = matrix[i-1][0].getScore(); |
---|
114 | maxRowIndex = 0; |
---|
115 | } else { |
---|
116 | tempMax = matrix[i-1][1].getScore(); |
---|
117 | maxRowIndex = 1; |
---|
118 | //position of the maximal score of the previous row |
---|
119 | |
---|
120 | for(int j = 2; j < length2;j++) { |
---|
121 | if(matrix[i-1][j].getScore() > tempMax) { |
---|
122 | tempMax = matrix[i-1][j].getScore(); |
---|
123 | maxRowIndex = j; |
---|
124 | } |
---|
125 | } |
---|
126 | |
---|
127 | } |
---|
128 | |
---|
129 | |
---|
130 | tempMax -= scoreThreshold; |
---|
131 | matrix[i][0].setScore(Math.max(firstRowLeftScore, tempMax)); |
---|
132 | if(tempMax ==matrix[i][0].getScore()){ |
---|
133 | matrix[i][0].setPrevious(matrix[i-1][maxRowIndex]); |
---|
134 | } |
---|
135 | |
---|
136 | if(firstRowLeftScore == matrix[i][0].getScore()) { |
---|
137 | matrix[i][0].setPrevious(matrix[i-1][0]); |
---|
138 | } |
---|
139 | |
---|
140 | //The last additional score is not related to a character in the input sequence, it's the total score. Therefore we don't need to save something for it |
---|
141 | if(i<length1+1) |
---|
142 | { |
---|
143 | matrix[i][0].setXvalue(input1[i-1]); |
---|
144 | matrix[i][0].setYvalue(-2); |
---|
145 | } |
---|
146 | else { |
---|
147 | //End after we calculated final score |
---|
148 | return; |
---|
149 | } |
---|
150 | |
---|
151 | |
---|
152 | for (int j = 1; j < length2; j++) { |
---|
153 | double diagScore = matrix[i - 1][j - 1].getScore() + similarity(i, j); |
---|
154 | double upScore = matrix[i][j - 1].getScore() + submat.getGapPenalty(); |
---|
155 | double leftScore = matrix[i - 1][j].getScore() + submat.getGapPenalty(); |
---|
156 | |
---|
157 | matrix[i][j].setScore(Math.max(diagScore,Math.max(upScore, Math.max(leftScore,matrix[i][0].getScore())))); |
---|
158 | |
---|
159 | // find the directions that give the maximum scores. |
---|
160 | // Multiple directions are ignored TODO |
---|
161 | //True if we had a match |
---|
162 | if (diagScore == matrix[i][j].getScore()) { |
---|
163 | matrix[i][j].setPrevious(matrix[i-1][j-1]); |
---|
164 | matrix[i][j].setXvalue(input1[i-1]); |
---|
165 | matrix[i][j].setYvalue(input2[j-1]); |
---|
166 | } |
---|
167 | //true if we took an event from sequence x and not from y |
---|
168 | if (leftScore == matrix[i][j].getScore()) { |
---|
169 | matrix[i][j].setXvalue(input1[i-1]); |
---|
170 | matrix[i][j].setYvalue(-1); |
---|
171 | matrix[i][j].setPrevious(matrix[i-1][j]); |
---|
172 | } |
---|
173 | //true if we took an event from sequence y and not from x |
---|
174 | if (upScore == matrix[i][j].getScore()) { |
---|
175 | matrix[i][j].setXvalue(-1); |
---|
176 | matrix[i][j].setYvalue(input2[j-1]); |
---|
177 | matrix[i][j].setPrevious(matrix[i][j-1]); |
---|
178 | } |
---|
179 | //true if we ended a matching region |
---|
180 | if (matrix[i][0].getScore() == matrix[i][j].getScore()) { |
---|
181 | matrix[i][j].setPrevious(matrix[i][0]); |
---|
182 | matrix[i][j].setXvalue(input1[i-1]); |
---|
183 | matrix[i][j].setYvalue(-2); |
---|
184 | } |
---|
185 | } |
---|
186 | |
---|
187 | //Set the complete score cell |
---|
188 | |
---|
189 | } |
---|
190 | } |
---|
191 | |
---|
192 | /** |
---|
193 | * Get the maximum value in the score matrix. |
---|
194 | */ |
---|
195 | public double getMaxScore() { |
---|
196 | double maxScore = 0; |
---|
197 | |
---|
198 | // skip the first row and column |
---|
199 | for (int i = 1; i <= length1; i++) { |
---|
200 | for (int j = 1; j < length2; j++) { |
---|
201 | if (matrix[i][j].getScore() > maxScore) { |
---|
202 | maxScore = matrix[i][j].getScore(); |
---|
203 | } |
---|
204 | } |
---|
205 | } |
---|
206 | |
---|
207 | return maxScore; |
---|
208 | } |
---|
209 | |
---|
210 | /** |
---|
211 | * Get the alignment score between the two input strings. |
---|
212 | */ |
---|
213 | public double getAlignmentScore() { |
---|
214 | return matrix[length1+1][0].getScore(); |
---|
215 | } |
---|
216 | |
---|
217 | |
---|
218 | |
---|
219 | public void traceback() { |
---|
220 | MatrixEntry tmp = matrix[length1+1][0]; |
---|
221 | //TODO: Why do we need such long arrays, also, add a check if we reach the limit? |
---|
222 | int aligned1[] = new int[2*length1+2*length2]; |
---|
223 | int aligned2[] = new int[2*length1+2*length2]; |
---|
224 | System.out.println("Aligned length: " + aligned1.length); |
---|
225 | int count = 0; |
---|
226 | do |
---|
227 | { |
---|
228 | if(count != 0) |
---|
229 | { |
---|
230 | aligned1[count] = tmp.getXvalue(); |
---|
231 | aligned2[count] = tmp.getYvalue(); |
---|
232 | } |
---|
233 | |
---|
234 | tmp = tmp.getPrevious(); |
---|
235 | count++; |
---|
236 | System.out.println(count); |
---|
237 | |
---|
238 | } while(tmp != null); |
---|
239 | |
---|
240 | //reverse order |
---|
241 | int reversed1[] = new int[count]; |
---|
242 | int reversed2[] = new int[count]; |
---|
243 | |
---|
244 | |
---|
245 | for(int i = count-1; i > 0; i--) { |
---|
246 | reversed1[reversed1.length-i]= aligned1[i]; |
---|
247 | reversed2[reversed2.length-i]= aligned2[i]; |
---|
248 | } |
---|
249 | |
---|
250 | NumberSequence ns1 = new NumberSequence(reversed1.length); |
---|
251 | NumberSequence ns2 = new NumberSequence(reversed2.length); |
---|
252 | ns1.setSequence(reversed1); |
---|
253 | ns2.setSequence(reversed2); |
---|
254 | |
---|
255 | alignment.add(ns1); |
---|
256 | alignment.add(ns2); |
---|
257 | } |
---|
258 | |
---|
259 | public void printAlignment() { |
---|
260 | MatrixEntry tmp = matrix[length1+1][0]; |
---|
261 | String aligned1 = ""; |
---|
262 | String aligned2 = ""; |
---|
263 | int count = 0; |
---|
264 | do |
---|
265 | { |
---|
266 | String append1=""; |
---|
267 | String append2=""; |
---|
268 | |
---|
269 | if(tmp.getXvalue() == -1) { |
---|
270 | append1 = " ___"; |
---|
271 | } |
---|
272 | else if(tmp.getXvalue() == -2) { |
---|
273 | append1 = " ..."; |
---|
274 | } |
---|
275 | else { |
---|
276 | append1 = String.format("%5d", tmp.getXvalue()); |
---|
277 | } |
---|
278 | |
---|
279 | if(tmp.getYvalue() == -1) { |
---|
280 | append2 = " ___"; |
---|
281 | } |
---|
282 | else if(tmp.getYvalue() == -2) { |
---|
283 | append2 = " ..."; |
---|
284 | } |
---|
285 | else { |
---|
286 | append2 = String.format("%5d", tmp.getYvalue()); |
---|
287 | } |
---|
288 | if(count != 0) |
---|
289 | { |
---|
290 | aligned1 = append1 + aligned1; |
---|
291 | aligned2 = append2 + aligned2; |
---|
292 | } |
---|
293 | |
---|
294 | tmp = tmp.getPrevious(); |
---|
295 | count++; |
---|
296 | |
---|
297 | } while(tmp != null); |
---|
298 | System.out.println(aligned1); |
---|
299 | System.out.println(aligned2); |
---|
300 | } |
---|
301 | |
---|
302 | |
---|
303 | |
---|
304 | /** |
---|
305 | * print the dynmaic programming matrix |
---|
306 | */ |
---|
307 | public void printDPMatrix() { |
---|
308 | System.out.print(" "); |
---|
309 | for (int i = 1; i <= length1; i++) |
---|
310 | System.out.format("%5d", input1[i - 1]); |
---|
311 | System.out.println(); |
---|
312 | for (int j = 0; j < length2; j++) { |
---|
313 | if (j > 0) |
---|
314 | System.out.format("%5d ",input2[j - 1]); |
---|
315 | else{ |
---|
316 | System.out.print(" "); |
---|
317 | } |
---|
318 | for (int i = 0; i <= length1 + 1; i++) { |
---|
319 | if((i<length1+1) || (i==length1+1 && j==0)) { |
---|
320 | System.out.format("%4.1f ",matrix[i][j].getScore()); |
---|
321 | } |
---|
322 | |
---|
323 | } |
---|
324 | System.out.println(); |
---|
325 | } |
---|
326 | } |
---|
327 | |
---|
328 | |
---|
329 | public List<NumberSequence> getAlignment() { |
---|
330 | return alignment; |
---|
331 | } |
---|
332 | |
---|
333 | public void setAlignment(List<NumberSequence> alignment) { |
---|
334 | this.alignment = alignment; |
---|
335 | } |
---|
336 | |
---|
337 | } |
---|