X-Git-Url: http://source.jalview.org/gitweb/?a=blobdiff_plain;f=src%2Fjalview%2Fanalysis%2Fscoremodels%2FScoreMatrix.java;h=b73f826e40a51c8f4fb745c7234fbfacb328261c;hb=dd97032c11b883017a9c8125ff30a58dfb2144bd;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..b73f826 100644 --- a/src/jalview/analysis/scoremodels/ScoreMatrix.java +++ b/src/jalview/analysis/scoremodels/ScoreMatrix.java @@ -20,94 +20,239 @@ */ package jalview.analysis.scoremodels; -import jalview.api.analysis.ScoreModelI; -import jalview.schemes.ResidueProperties; +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; -public class ScoreMatrix extends PairwiseSeqScoreModel implements - ScoreModelI +import java.util.Arrays; + +public class ScoreMatrix implements SimilarityScoreModelI, + PairwiseScoreModelI { - String name; + private static final char GAP_CHARACTER = Comparison.GAP_DASH; - @Override - public String getName() - { - return name; - } + /* + * 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; - /** - * reference to integer score matrix + 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 */ - int[][] matrix; + private String name; - /** - * 0 for Protein Score matrix. 1 for dna score matrix + /* + * 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 */ - int type; + private boolean peptide; /** + * 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 matrix - * Pairwise scores indexed according to appropriate symbol alphabet - * @param type - * 0 for Protein, 1 for NA + * @param alphabet + * the symbols to which scores apply + * @param values + * Pairwise scores indexed according to the symbol alphabet */ - public ScoreMatrix(String name, int[][] matrix, int type) + public ScoreMatrix(String theName, char[] alphabet, float[][] values) { - this.matrix = matrix; - this.type = type; - 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); + + /* + * crude heuristic for now... + */ + peptide = alphabet.length >= 20; + } + + /** + * 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). + * + * @param alphabet + * @return + */ + static 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() + { + 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, or zero if c or d is + * an unscored or unexpected character + */ @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]; - - pog = matrix[a][b]; - } catch (Exception e) + System.err.println(String.format(BAD_ASCII_ERROR, c)); + return 0; + } + if (d >= symbolIndex.length) { - // System.out.println("Unknown residue in " + A1 + " " + A2); + System.err.println(String.format(BAD_ASCII_ERROR, d)); + return 0; } - return pog; + int cIndex = symbolIndex[c]; + int dIndex = symbolIndex[d]; + if (cIndex != UNMAPPED && dIndex != UNMAPPED) + { + return matrix[cIndex][dIndex]; + } + return 0; } /** @@ -119,57 +264,244 @@ 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(""); + sb.append(html ? "" : ""); + } + else + { + sb.append("ScoreMatrix ").append(getName()).append("\n"); + sb.append(symbols).append("\n"); + } + for (char sym : symbols) + { + if (html) + { + sb.append(""); + } + else + { + sb.append("\t").append(sym); + } + } + sb.append(html ? "\n" : "\n"); + + /* + * table of scores + */ + for (char c1 : symbols) + { + if (html) + { + sb.append("" : ""); + for (char c2 : symbols) + { + sb.append(html ? "" : ""); + } + sb.append(html ? "\n" : "\n"); + } + if (html) + { + sb.append("
 ").append(sym).append(" 
"); + } + sb.append(c1).append(html ? "" : "\t") + .append(matrix[symbolIndex[c1]][symbolIndex[c2]]) + .append(html ? "
"); } - for (char sym = 'A'; sym <= 'Z'; sym++) + 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 + *

+ */ + @Override + public MatrixI findSimilarities(AlignmentView seqstrings, + SimilarityParamsI options) + { + char gapChar = scoreGapAsAny ? (seqstrings.isNa() ? 'N' : 'X') + : Comparison.GAP_DASH; + String[] seqs = seqstrings.getSequenceStrings(gapChar); + return findSimilarities(seqs, options); + } + + /** + * Computes pairwise similarities of a set of sequences using the given + * parameters + * + * @param seqs + * @param params + * @return + */ + protected MatrixI findSimilarities(String[] seqs, SimilarityParamsI params) + { + double[][] values = new double[seqs.length][]; + for (int row = 0; row < seqs.length; row++) { - if (symbols[sym] >= 0 && symbols[sym] < symMax) + values[row] = new double[seqs.length]; + for (int col = 0; col < seqs.length; col++) { - if (header) + 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) + { + /* + * off the end of one sequence; stop if we are only matching + * on the shorter sequence length, else treat as trailing gap + */ + if (params.denominateByShortestLength()) { - sb.append(html ? "" : ""); - for (char sym2 = 'A'; sym2 <= 'Z'; sym2++) - { - if (symbols[sym2] >= 0 && symbols[sym2] < symMax) - { - sb.append((html ? " " : "\t") + sym2 - + (html ? " " : "")); - } - } - header = false; - sb.append(html ? "\n" : "\n"); + break; } - if (html) + } + + char c1 = i >= len1 ? GAP_CHARACTER : seq1.charAt(i); + char c2 = i >= len2 ? GAP_CHARACTER : seq2.charAt(i); + boolean gap1 = Comparison.isGap(c1); + boolean gap2 = Comparison.isGap(c2); + + if (gap1 && gap2) + { + /* + * gap-gap: include if options say so, else ignore + */ + if (!params.includeGappedColumns()) { - sb.append(""); + continue; } - sb.append((html ? "" : "") + sym + (html ? "" : "")); - for (char sym2 = 'A'; sym2 <= 'Z'; sym2++) + } + else if (gap1 || gap2) + { + /* + * gap-residue: score if options say so + */ + if (!params.includeGaps()) { - if (symbols[sym2] >= 0 && symbols[sym2] < symMax) - { - sb.append((html ? "" : "\t") - + matrix[symbols[sym]][symbols[sym2]] - + (html ? "" : "")); - } + continue; } - sb.append(html ? "\n" : "\n"); } + float score = getPairwiseScore(c1, c2); + total += score; } - if (html) + 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) { - sb.append(""); + hs = hs * 31 + Arrays.hashCode(row); } - return sb.toString(); + 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; + } + + /** + * 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; } }