*/
package jalview.analysis.scoremodels;
-import jalview.api.analysis.ScoreModelI;
+import jalview.math.Matrix;
+import jalview.math.MatrixI;
import java.util.Arrays;
-public class ScoreMatrix extends PairwiseSeqScoreModel implements
- ScoreModelI
+public class ScoreMatrix implements PairwiseScoreModelI
{
public static final short UNMAPPED = (short) -1;
private boolean peptide;
/**
- * Constructor
+ * Constructor given a name, symbol alphabet, and matrix of scores for pairs
+ * of symbols. The matrix should be square and of the same size as the
+ * alphabet, for example 20x20 for a 20 symbol alphabet.
*
* @param name
* Unique, human readable name for the matrix
*/
public ScoreMatrix(String name, char[] alphabet, float[][] matrix)
{
+ if (alphabet.length != matrix.length)
+ {
+ throw new IllegalArgumentException(
+ "score matrix size must match alphabet size");
+ }
+ for (float[] row : matrix)
+ {
+ if (row.length != alphabet.length)
+ {
+ throw new IllegalArgumentException(
+ "score matrix size must be square");
+ }
+ }
+
this.matrix = matrix;
this.name = name;
this.symbols = alphabet;
return peptide;
}
- @Override
+ /**
+ * Returns the score matrix as used in getPairwiseScore. If using this matrix
+ * directly, callers <em>must</em> also call <code>getMatrixIndex</code> in
+ * order to get the matrix index for each character (symbol).
+ *
+ * @return
+ * @see #getMatrixIndex(char)
+ */
public float[][] getMatrix()
{
return matrix;
}
/**
+ * Answers the matrix index for a given character, or -1 if unmapped in the
+ * matrix. Use this method only if using <code>getMatrix</code> in order to
+ * compute scores directly (without symbol lookup) for efficiency.
+ *
+ * @param c
+ * @return
+ * @see #getMatrix()
+ */
+ public int getMatrixIndex(char c)
+ {
+ if (c < symbolIndex.length)
+ {
+ return symbolIndex[c];
+ }
+ else
+ {
+ return UNMAPPED;
+ }
+ }
+
+ /**
* Returns the pairwise score for substituting c with d, or zero if c or d is
* an unscored or unexpected character
*/
@Override
public float getPairwiseScore(char c, char d)
{
- if (c > MAX_ASCII)
+ if (c >= symbolIndex.length)
{
System.err.println(String.format(BAD_ASCII_ERROR, c));
return 0;
}
- if (d > MAX_ASCII)
+ if (d >= symbolIndex.length)
{
System.err.println(String.format(BAD_ASCII_ERROR, d));
return 0;
}
/**
- * Print the score matrix, optionally formatted as html, with the alphabet symbols as column headings and at the start of each row
+ * Print the score matrix, optionally formatted as html, with the alphabet
+ * symbols as column headings and at the start of each row.
+ * <p>
+ * The non-html format should give an output which can be parsed as a score
+ * matrix file
+ *
* @param html
* @return
*/
sb.append("<table border=\"1\">");
sb.append(html ? "<tr><th></th>" : "");
}
+ else
+ {
+ sb.append("ScoreMatrix ").append(getName()).append("\n");
+ sb.append(symbols).append("\n");
+ }
for (char sym : symbols)
{
if (html)
}
return sb.toString();
}
+
+ /**
+ * Answers the number of symbols coded for (also equal to the number of rows
+ * and columns of the score matrix)
+ *
+ * @return
+ */
+ public int getSize()
+ {
+ return symbols.length;
+ }
+
+ /**
+ * Computes an NxN matrix where N is the number of sequences, and entry [i, j]
+ * is sequence[i] pairwise multiplied with sequence[j], as a sum of scores
+ * computed using the current score matrix. For example
+ * <ul>
+ * <li>Sequences:</li>
+ * <li>FKL</li>
+ * <li>R-D</li>
+ * <li>QIA</li>
+ * <li>GWC</li>
+ * <li>Score matrix is BLOSUM62</li>
+ * <li>Gaps treated same as X (unknown)</li>
+ * <li>product [0, 0] = F.F + K.K + L.L = 6 + 5 + 4 = 15</li>
+ * <li>product [1, 1] = R.R + -.- + D.D = 5 + -1 + 6 = 10</li>
+ * <li>product [2, 2] = Q.Q + I.I + A.A = 5 + 4 + 4 = 13</li>
+ * <li>product [3, 3] = G.G + W.W + C.C = 6 + 11 + 9 = 26</li>
+ * <li>product[0, 1] = F.R + K.- + L.D = -3 + -1 + -3 = -8
+ * <li>and so on</li>
+ * </ul>
+ */
+ public MatrixI computePairwiseScores(String[] seqs)
+ {
+ double[][] values = new double[seqs.length][];
+ for (int row = 0; row < seqs.length; row++)
+ {
+ values[row] = new double[seqs.length];
+ for (int col = 0; col < seqs.length; col++)
+ {
+ int total = 0;
+ int width = Math.min(seqs[row].length(), seqs[col].length());
+ for (int i = 0; i < width; i++)
+ {
+ char c1 = seqs[row].charAt(i);
+ char c2 = seqs[col].charAt(i);
+ float score = getPairwiseScore(c1, c2);
+ total += score;
+ }
+ values[row][col] = total;
+ }
+ }
+ return new Matrix(values);
+ }
+
+ /**
+ * Answers a hashcode computed from the symbol alphabet and the matrix score
+ * values
+ */
+ @Override
+ public int hashCode()
+ {
+ int hs = Arrays.hashCode(symbols);
+ for (float[] row : matrix)
+ {
+ hs = hs * 31 + Arrays.hashCode(row);
+ }
+ return hs;
+ }
+
+ /**
+ * Answers true if the argument is a ScoreMatrix with the same symbol alphabet
+ * and score values, else false
+ */
+ @Override
+ public boolean equals(Object obj)
+ {
+ if (!(obj instanceof ScoreMatrix))
+ {
+ return false;
+ }
+ ScoreMatrix sm = (ScoreMatrix) obj;
+ if (Arrays.equals(symbols, sm.symbols)
+ && Arrays.deepEquals(matrix, sm.matrix))
+ {
+ return true;
+ }
+ return false;
+ }
}