X-Git-Url: http://source.jalview.org/gitweb/?a=blobdiff_plain;f=src%2Fjalview%2Fanalysis%2Fscoremodels%2FScoreMatrix.java;h=b206339805cf4d1f026c15eb34f3987a21a543b9;hb=f5e62b7ae1c880870d2c162bcd3aa3097f799b41;hp=41aef8232fad754aeb6b6febfefc14483fa3881a;hpb=1750b2f1dcdb6e7de41d0ef7beda88dc3400afba;p=jalview.git diff --git a/src/jalview/analysis/scoremodels/ScoreMatrix.java b/src/jalview/analysis/scoremodels/ScoreMatrix.java index 41aef82..b206339 100644 --- a/src/jalview/analysis/scoremodels/ScoreMatrix.java +++ b/src/jalview/analysis/scoremodels/ScoreMatrix.java @@ -20,94 +20,340 @@ */ package jalview.analysis.scoremodels; +import jalview.api.AlignmentViewPanel; +import jalview.api.analysis.PairwiseScoreModelI; import jalview.api.analysis.ScoreModelI; -import jalview.schemes.ResidueProperties; +import jalview.api.analysis.SimilarityParamsI; +import jalview.datamodel.AlignmentView; +import jalview.math.Matrix; +import jalview.math.MatrixI; +import jalview.util.Comparison; -public class ScoreMatrix extends PairwiseSeqScoreModel implements - ScoreModelI +import java.util.Arrays; + +/** + * A class that models a substitution score matrix for any given alphabet of + * symbols. Instances of this class are immutable and thread-safe, so the same + * object is returned from calls to getInstance(). + */ +public class ScoreMatrix extends SimilarityScoreModel + implements PairwiseScoreModelI { - String name; + private static final char GAP_CHARACTER = Comparison.GAP_DASH; - @Override - public String getName() + /* + * an arbitrary score to assign for identity of an unknown symbol + * (this is the value on the diagonal in the * column of the NCBI matrix) + * (though a case could be made for using the minimum diagonal value) + */ + private static final int UNKNOWN_IDENTITY_SCORE = 1; + + /* + * Jalview 2.10.1 treated gaps as X (peptide) or N (nucleotide) + * for pairwise scoring; 2.10.2 uses gap score (last column) in + * score matrix (JAL-2397) + * Set this flag to true (via Groovy) for 2.10.1 behaviour + */ + private static boolean scoreGapAsAny = false; + + public static final short UNMAPPED = (short) -1; + + private static final String BAD_ASCII_ERROR = "Unexpected character %s in getPairwiseScore"; + + private static final int MAX_ASCII = 127; + + /* + * the name of the model as shown in menus + * each score model in use should have a unique name + */ + private String name; + + /* + * a description for the model as shown in tooltips + */ + private String description; + + /* + * the characters that the model provides scores for + */ + private char[] symbols; + + /* + * the score matrix; both dimensions must equal the number of symbols + * matrix[i][j] is the substitution score for replacing symbols[i] with symbols[j] + */ + private float[][] matrix; + + /* + * quick lookup to convert from an ascii character value to the index + * of the corresponding symbol in the score matrix + */ + private short[] symbolIndex; + + /* + * true for Protein Score matrix, false for dna score matrix + */ + private boolean peptide; + + private float minValue; + + private float maxValue; + + private boolean symmetric; + + /** + * 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 theName + * Unique, human readable name for the matrix + * @param alphabet + * the symbols to which scores apply + * @param values + * Pairwise scores indexed according to the symbol alphabet + */ + public ScoreMatrix(String theName, char[] alphabet, float[][] values) { - return name; + this(theName, null, alphabet, values); + } + + /** + * Constructor given a name, description, 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 theName + * Unique, human readable name for the matrix + * @param theDescription + * descriptive display name suitable for use in menus + * @param alphabet + * the symbols to which scores apply + * @param values + * Pairwise scores indexed according to the symbol alphabet + */ + public ScoreMatrix(String theName, String theDescription, char[] alphabet, + float[][] values) + { + if (alphabet.length != values.length) + { + throw new IllegalArgumentException( + "score matrix size must match alphabet size"); + } + for (float[] row : values) + { + if (row.length != alphabet.length) + { + throw new IllegalArgumentException( + "score matrix size must be square"); + } + } + + this.matrix = values; + this.name = theName; + this.description = theDescription; + this.symbols = alphabet; + + symbolIndex = buildSymbolIndex(alphabet); + + findMinMax(); + + symmetric = checkSymmetry(); + + /* + * crude heuristic for now... + */ + peptide = alphabet.length >= 20; } /** - * reference to integer score matrix + * Answers true if the matrix is symmetric, else false. Usually, substitution + * matrices are symmetric, which allows calculations to be short cut. + * + * @return */ - int[][] matrix; + private boolean checkSymmetry() + { + for (int i = 0; i < matrix.length; i++) + { + for (int j = i; j < matrix.length; j++) + { + if (matrix[i][j] != matrix[j][i]) + { + return false; + } + } + } + return true; + } /** - * 0 for Protein Score matrix. 1 for dna score matrix + * Record the minimum and maximum score values */ - int type; + protected void findMinMax() + { + float min = Float.MAX_VALUE; + float max = -Float.MAX_VALUE; + if (matrix != null) + { + for (float[] row : matrix) + { + if (row != null) + { + for (float f : row) + { + min = Math.min(min, f); + max = Math.max(max, f); + } + } + } + } + minValue = min; + maxValue = max; + } /** + * Returns an array A where A[i] is the position in the alphabet array of the + * character whose value is i. For example if the alphabet is { 'A', 'D', 'X' + * } then A['D'] = A[68] = 1. + *
+ * Unmapped characters (not in the alphabet) get an index of -1. + *
+ * Mappings are added automatically for lower case symbols (for non case + * sensitive scoring), unless they are explicitly present in the alphabet (are + * scored separately in the score matrix). + *
+ * the gap character (space, dash or dot) included in the alphabet (if any) is
+ * recorded in a field
*
- * @param name
- * Unique, human readable name for the matrix
- * @param matrix
- * Pairwise scores indexed according to appropriate symbol alphabet
- * @param type
- * 0 for Protein, 1 for NA
+ * @param alphabet
+ * @return
*/
- public ScoreMatrix(String name, int[][] matrix, int type)
+ short[] buildSymbolIndex(char[] alphabet)
+ {
+ short[] index = new short[MAX_ASCII + 1];
+ Arrays.fill(index, UNMAPPED);
+ short pos = 0;
+ for (char c : alphabet)
+ {
+ if (c <= MAX_ASCII)
+ {
+ index[c] = pos;
+ }
+
+ /*
+ * also map lower-case character (unless separately mapped)
+ */
+ if (c >= 'A' && c <= 'Z')
+ {
+ short lowerCase = (short) (c + ('a' - 'A'));
+ if (index[lowerCase] == UNMAPPED)
+ {
+ index[lowerCase] = pos;
+ }
+ }
+ pos++;
+ }
+ return index;
+ }
+
+ @Override
+ public String getName()
{
- this.matrix = matrix;
- this.type = type;
- this.name = name;
+ return name;
+ }
+
+ @Override
+ public String getDescription()
+ {
+ return description;
}
@Override
public boolean isDNA()
{
- return type == 1;
+ return !peptide;
}
@Override
public boolean isProtein()
{
- return type == 0;
+ return peptide;
}
- @Override
- public int[][] getMatrix()
+ /**
+ * Returns a copy of the score matrix as used in getPairwiseScore. If using
+ * this matrix directly, callers must also call
+ * getMatrixIndex
in order to get the matrix index for each
+ * character (symbol).
+ *
+ * @return
+ * @see #getMatrixIndex(char)
+ */
+ public float[][] getMatrix()
{
- return matrix;
+ float[][] v = new float[matrix.length][matrix.length];
+ for (int i = 0; i < matrix.length; i++)
+ {
+ v[i] = Arrays.copyOf(matrix[i], matrix[i].length);
+ }
+ return v;
}
/**
+ * Answers the matrix index for a given character, or -1 if unmapped in the
+ * matrix. Use this method only if using getMatrix
in order to
+ * compute scores directly (without symbol lookup) for efficiency.
*
- * @param A1
- * @param A2
- * @return score for substituting first char in A1 with first char in A2
+ * @param c
+ * @return
+ * @see #getMatrix()
*/
- public int getPairwiseScore(String A1, String A2)
+ public int getMatrixIndex(char c)
{
- return getPairwiseScore(A1.charAt(0), A2.charAt(0));
+ if (c < symbolIndex.length)
+ {
+ return symbolIndex[c];
+ }
+ else
+ {
+ return UNMAPPED;
+ }
}
+ /**
+ * Returns the pairwise score for substituting c with d. If either c or d is
+ * an unexpected character, returns 1 for identity (c == d), else the minimum
+ * score value in the matrix.
+ */
@Override
- public int getPairwiseScore(char c, char d)
+ public float getPairwiseScore(char c, char d)
{
- int pog = 0;
-
- try
+ if (c >= symbolIndex.length)
{
- int a = (type == 0) ? ResidueProperties.aaIndex[c]
- : ResidueProperties.nucleotideIndex[c];
- int b = (type == 0) ? ResidueProperties.aaIndex[d]
- : ResidueProperties.nucleotideIndex[d];
+ System.err.println(String.format(BAD_ASCII_ERROR, c));
+ return 0;
+ }
+ if (d >= symbolIndex.length)
+ {
+ System.err.println(String.format(BAD_ASCII_ERROR, d));
+ return 0;
+ }
- pog = matrix[a][b];
- } catch (Exception e)
+ int cIndex = symbolIndex[c];
+ int dIndex = symbolIndex[d];
+ if (cIndex != UNMAPPED && dIndex != UNMAPPED)
{
- // System.out.println("Unknown residue in " + A1 + " " + A2);
+ return matrix[cIndex][dIndex];
}
- return pog;
+ /*
+ * one or both symbols not found in the matrix
+ * currently scoring as 1 (for identity) or the minimum
+ * matrix score value (otherwise)
+ * (a case could be made for using minimum row/column value instead)
+ */
+ return c == d ? UNKNOWN_IDENTITY_SCORE : getMinimumScore();
}
/**
@@ -119,57 +365,264 @@ public class ScoreMatrix extends PairwiseSeqScoreModel implements
return outputMatrix(false);
}
+ /**
+ * Print the score matrix, optionally formatted as html, with the alphabet
+ * symbols as column headings and at the start of each row.
+ *
+ * The non-html format should give an output which can be parsed as a score + * matrix file + * + * @param html + * @return + */ public String outputMatrix(boolean html) { - StringBuffer sb = new StringBuffer(); - int[] symbols = (type == 0) ? ResidueProperties.aaIndex - : ResidueProperties.nucleotideIndex; - int symMax = (type == 0) ? ResidueProperties.maxProteinIndex - : ResidueProperties.maxNucleotideIndex; - boolean header = true; + StringBuilder sb = new StringBuilder(512); + + /* + * heading row with alphabet + */ if (html) { sb.append("
" : ""); + } + else + { + sb.append("ScoreMatrix ").append(getName()).append("\n"); + } + for (char sym : symbols) + { + if (html) + { + sb.append(" | ").append(sym).append(" | "); + } + else + { + sb.append("\t").append(sym); + } + } + sb.append(html ? "
---|---|
"); + } + sb.append(c1).append(html ? " | " : ""); + for (char c2 : symbols) + { + sb.append(html ? "" : "\t") + .append(matrix[symbolIndex[c1]][symbolIndex[c2]]) + .append(html ? " | " : ""); + } + sb.append(html ? "