*/
package jalview.analysis.scoremodels;
-import jalview.api.analysis.ScoreModelI;
+import jalview.api.analysis.PairwiseScoreModelI;
+import jalview.api.analysis.SimilarityParamsI;
+import jalview.api.analysis.SimilarityScoreModelI;
+import jalview.datamodel.AlignmentView;
+import jalview.math.Matrix;
+import jalview.math.MatrixI;
+import jalview.util.Comparison;
-import java.io.BufferedReader;
-import java.io.IOException;
-import java.io.InputStream;
-import java.io.InputStreamReader;
import java.util.Arrays;
-import java.util.StringTokenizer;
-public class ScoreMatrix extends PairwiseSeqScoreModel implements
- ScoreModelI
+/**
+ * A class that models a substitution score matrix for any given alphabet of
+ * symbols
+ */
+public class ScoreMatrix implements SimilarityScoreModelI,
+ PairwiseScoreModelI
{
- public static final short UNMAPPED = (short) -1;
+ /*
+ * this fields records which gap character (if any) is used in the alphabet;
+ * space, dash or dot are recognised as gap symbols
+ */
+ private char gapCharacter = '0';
- private static final String DELIMITERS = " ,\t";
+ /*
+ * 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;
- private static final String COMMENT_CHAR = "#";
+ public static final short UNMAPPED = (short) -1;
private static final String BAD_ASCII_ERROR = "Unexpected character %s in getPairwiseScore";
/*
* 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;
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
+ * @param theName
* Unique, human readable name for the matrix
* @param alphabet
* the symbols to which scores apply
- * @param matrix
+ * @param values
* Pairwise scores indexed according to the symbol alphabet
*/
- public ScoreMatrix(String name, char[] alphabet, float[][] matrix)
+ public ScoreMatrix(String theName, char[] alphabet, float[][] values)
{
- this.matrix = matrix;
- this.name = name;
+ 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.symbols = alphabet;
symbolIndex = buildSymbolIndex(alphabet);
* 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).
+ * <p>
+ * the gap character (space, dash or dot) included in the alphabet (if any) is
+ * recorded in a field
*
* @param alphabet
* @return
*/
- static short[] buildSymbolIndex(char[] alphabet)
+ short[] buildSymbolIndex(char[] alphabet)
{
short[] index = new short[MAX_ASCII + 1];
Arrays.fill(index, UNMAPPED);
short pos = 0;
for (char c : alphabet)
{
+ if (Comparison.isGap(c))
+ {
+ gapCharacter = c;
+ }
+
if (c <= MAX_ASCII)
{
index[c] = pos;
}
@Override
+ public String getDescription()
+ {
+ return description;
+ }
+
+ @Override
public boolean isDNA()
{
return !peptide;
return peptide;
}
- @Override
+ /**
+ * Returns a copy of 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;
+ 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 <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;
+ }
+ }
+
+ /**
+ * Answers the matrix index for the gap 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.
+ *
+ * @return
+ * @see #getMatrix()
+ */
+ public int getGapIndex()
+ {
+ return getMatrixIndex(gapCharacter);
}
/**
@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)
}
/**
- * Parse a score matrix from the given input stream and returns a ScoreMatrix
- * object. If parsing fails, error messages are written to syserr and null is
- * returned. It is the caller's responsibility to close the input stream.
- * Expected format:
+ * Answers the number of symbols coded for (also equal to the number of rows
+ * and columns of the score matrix)
*
- * <pre>
- * ScoreMatrix displayName
- * # comment lines begin with hash sign
- * # symbol alphabet should be the next non-comment line
- * ARNDCQEGHILKMFPSTWYVBZX *
- * # scores matrix, with space, comma or tab delimited values
- * # [i, j] = score for substituting symbol[i] with symbol[j]
- * # first column in each row is optionally the 'substituted' symbol
- * A 4 -1 -2 -2 0 -1 -1 0 -2 -1 -1 -1 -1 -2 -1 1 0 -3 -2 0 -2 -1 0 -4 -4
- * ..etc..
- * </pre>
+ * @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>
+ */
+ @Override
+ public MatrixI findSimilarities(AlignmentView seqstrings,
+ SimilarityParamsI options)
+ {
+ char gapChar = scoreGapAsAny ? (seqstrings.isNa() ? 'N' : 'X')
+ : gapCharacter;
+ String[] seqs = seqstrings.getSequenceStrings(gapChar);
+ return findSimilarities(seqs, options);
+ }
+
+ /**
+ * Computes pairwise similarities of a set of sequences using the given
+ * parameters
*
- * @param is
+ * @param seqs
+ * @param params
* @return
*/
- public static ScoreMatrix parse(InputStream is)
+ protected MatrixI findSimilarities(String[] seqs, SimilarityParamsI params)
{
- ScoreMatrix sm = null;
- BufferedReader br = new BufferedReader(new InputStreamReader(is));
- int lineNo = 0;
- String name = null;
- String alphabet = null;
- float[][] scores = null;
- int size = 0;
- int row = 0;
-
- try
+ double[][] values = new double[seqs.length][];
+ for (int row = 0; row < seqs.length; row++)
{
- String data;
-
- while ((data = br.readLine()) != null)
+ values[row] = new double[seqs.length];
+ for (int col = 0; col < seqs.length; col++)
{
- lineNo++;
- data = data.trim();
- if (data.startsWith(COMMENT_CHAR))
- {
- continue;
- }
- if (data.toLowerCase().startsWith("scorematrix"))
- {
- /*
- * Parse name from ScoreMatrix <name>
- */
- if (name != null)
- {
- System.err
- .println("Warning: 'ScoreMatrix' repeated in file at line "
- + lineNo);
- }
- StringTokenizer nameLine = new StringTokenizer(data, DELIMITERS);
- if (nameLine.countTokens() != 2)
- {
- System.err
- .println("Format error: expected 'ScoreMatrix <name>', found '"
- + data + "' at line " + lineNo);
- return null;
- }
- nameLine.nextToken();
- name = nameLine.nextToken();
- continue;
- }
- else if (name == null)
- {
- System.err
- .println("Format error: 'ScoreMatrix <name>' should be the first non-comment line");
- return null;
- }
+ double total = computeSimilarity(seqs[row], seqs[col], params);
+ values[row][col] = total;
+ }
+ }
+ return new Matrix(values);
+ }
+
+ /**
+ * Calculates the pairwise similarity of two strings using the given
+ * calculation parameters
+ *
+ * @param seq1
+ * @param seq2
+ * @param params
+ * @return
+ */
+ protected double computeSimilarity(String seq1, String seq2,
+ SimilarityParamsI params)
+ {
+ int len1 = seq1.length();
+ int len2 = seq2.length();
+ double total = 0;
+ int width = Math.max(len1, len2);
+ for (int i = 0; i < width; i++)
+ {
+ if (i >= len1 || i >= len2)
+ {
/*
- * next line after ScoreMatrix should be the alphabet of scored symbols
+ * off the end of one sequence; stop if we are only matching
+ * on the shorter sequence length, else treat as trailing gap
*/
- if (alphabet == null)
+ if (params.denominateByShortestLength())
{
- alphabet = data;
- size = alphabet.length();
- scores = new float[size][];
- continue;
+ break;
}
+ }
+
+ char c1 = i >= len1 ? gapCharacter : seq1.charAt(i);
+ char c2 = i >= len2 ? gapCharacter : seq2.charAt(i);
+ boolean gap1 = Comparison.isGap(c1);
+ boolean gap2 = Comparison.isGap(c2);
+ if (gap1 && gap2)
+ {
/*
- * too much information?
+ * gap-gap: include if options say so, else ignore
*/
- if (row >= size && data.length() > 0) {
- System.err
- .println("Unexpected extra input line in score model file "
- + data);
- return null;
+ if (!params.includeGappedColumns())
+ {
+ continue;
}
-
+ }
+ else if (gap1 || gap2)
+ {
/*
- * subsequent lines should be the symbol scores
- * optionally with the symbol as the first column for readability
+ * gap-residue: score if options say so
*/
- StringTokenizer scoreLine = new StringTokenizer(data, DELIMITERS);
- if (scoreLine.countTokens() == size + 1)
+ if (!params.includeGaps())
{
- /*
- * check 'guide' symbol is the row'th letter of the alphabet
- */
- String symbol = scoreLine.nextToken();
- if (symbol.length() > 1
- || symbol.charAt(0) != alphabet.charAt(row))
- {
- System.err
- .println(String
- .format("Error parsing score matrix at line %d, expected %s but found %s",
- lineNo, alphabet.charAt(row), symbol));
- return null;
- }
- }
- if (scoreLine.countTokens() != size)
- {
- System.err.println(String.format(
- "Expected %d scores at line %d but found %d", size,
- lineNo, scoreLine.countTokens()));
- return null;
- }
- scores[row] = new float[size];
- int col = 0;
- String value = null;
- while (scoreLine.hasMoreTokens())
- {
- try
- {
- value = scoreLine.nextToken();
- scores[row][col] = Float.valueOf(value);
- col++;
- } catch (NumberFormatException e)
- {
- System.err.println(String.format(
- "Invalid score value %s at line %d column %d", value,
- lineNo, col));
- return null;
- }
+ continue;
}
- row++;
}
- } catch (IOException e)
+ float score = getPairwiseScore(c1, c2);
+ total += score;
+ }
+ return total;
+ }
+
+ /**
+ * 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)
{
- System.err.println("Error reading score matrix file: "
- + e.getMessage() + " at line " + lineNo);
+ hs = hs * 31 + Arrays.hashCode(row);
}
+ return hs;
+ }
- /*
- * out of data - check we found enough
- */
- if (row < size)
+ /**
+ * 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))
{
- System.err
- .println(String
- .format("Expected %d rows of score data in score matrix but only found %d",
- size, row));
- return null;
+ return true;
}
+ return false;
+ }
- /*
- * If we get here, then name, alphabet and scores have been parsed successfully
- */
- sm = new ScoreMatrix(name, alphabet.toCharArray(), scores);
- return sm;
+ /**
+ * Returns the alphabet the matrix scores for, as a string of characters
+ *
+ * @return
+ */
+ public String getSymbols()
+ {
+ return new String(symbols);
+ }
+
+ public void setDescription(String desc)
+ {
+ description = desc;
}
}
git://source.jalview.org / jalview.git / blobdiff