X-Git-Url: http://source.jalview.org/gitweb/?a=blobdiff_plain;f=src%2Fjalview%2Fanalysis%2FAAFrequency.java;h=1fef08ecbc8e8014ca392d86f171816a494d8882;hb=da74fb8057a0eb3071da3b8dd29d0381a4d2aa00;hp=50045dc72a620ce73b344003c526c657109f0611;hpb=fddf3084802b37e5cee17829e32692a4aac3e60d;p=jalview.git diff --git a/src/jalview/analysis/AAFrequency.java b/src/jalview/analysis/AAFrequency.java index 50045dc..1fef08e 100755 --- a/src/jalview/analysis/AAFrequency.java +++ b/src/jalview/analysis/AAFrequency.java @@ -24,15 +24,25 @@ import jalview.datamodel.AlignedCodonFrame; import jalview.datamodel.AlignmentAnnotation; import jalview.datamodel.AlignmentI; import jalview.datamodel.Annotation; +import jalview.datamodel.HiddenMarkovModel; +import jalview.datamodel.Profile; +import jalview.datamodel.ProfileI; +import jalview.datamodel.Profiles; +import jalview.datamodel.ProfilesI; +import jalview.datamodel.ResidueCount; +import jalview.datamodel.ResidueCount.SymbolCounts; import jalview.datamodel.SequenceI; +import jalview.ext.android.SparseIntArray; +import jalview.schemes.ResidueProperties; +import jalview.util.Comparison; import jalview.util.Format; import jalview.util.MappingUtils; import jalview.util.QuickSort; +import java.awt.Color; import java.util.Arrays; import java.util.Hashtable; import java.util.List; -import java.util.Set; /** * Takes in a vector or array of sequences and column start and column end and @@ -40,45 +50,20 @@ import java.util.Set; * This class is used extensively in calculating alignment colourschemes that * depend on the amount of conservation in each alignment column. * - * @author $author$ - * @version $Revision$ */ public class AAFrequency { - private static final int TO_UPPER_CASE = 'A' - 'a'; // -32 - - public static final String MAXCOUNT = "C"; - - public static final String MAXRESIDUE = "R"; - - public static final String PID_GAPS = "G"; - - public static final String PID_NOGAPS = "N"; + private static final double LOG2 = Math.log(2); public static final String PROFILE = "P"; - public static final String ENCODED_CHARS = "E"; - - /* - * Quick look-up of String value of char 'A' to 'Z' - */ - private static final String[] CHARS = new String['Z' - 'A' + 1]; - - static - { - for (char c = 'A'; c <= 'Z'; c++) - { - CHARS[c - 'A'] = String.valueOf(c); - } - } - - public static final Hashtable[] calculate(List list, - int start, int end) + public static final ProfilesI calculate(List list, int start, + int end) { return calculate(list, start, end, false); } - public static final Hashtable[] calculate(List sequences, + public static final ProfilesI calculate(List sequences, int start, int end, boolean profile) { SequenceI[] seqs = new SequenceI[sequences.size()]; @@ -88,309 +73,431 @@ public class AAFrequency for (int i = 0; i < sequences.size(); i++) { seqs[i] = sequences.get(i); - if (seqs[i].getLength() > width) + int length = seqs[i].getLength(); + if (length > width) { - width = seqs[i].getLength(); + width = length; } } - Hashtable[] reply = new Hashtable[width]; - if (end >= width) { end = width; } - calculate(seqs, start, end, reply, profile); + ProfilesI reply = calculate(seqs, width, start, end, profile); return reply; } } - public static final void calculate(SequenceI[] sequences, int start, - int end, Hashtable[] result, boolean profile) + /** + * Calculate the consensus symbol(s) for each column in the given range. + * + * @param sequences + * @param width + * the full width of the alignment + * @param start + * start column (inclusive, base zero) + * @param end + * end column (exclusive) + * @param saveFullProfile + * if true, store all symbol counts + */ + public static final ProfilesI calculate(final SequenceI[] sequences, + int width, int start, int end, boolean saveFullProfile) { - Hashtable residueHash; - int maxCount, nongap, i, j, v; - int jSize = sequences.length; - String maxResidue; - char c = '-'; - float percentage; + // long now = System.currentTimeMillis(); + int seqCount = sequences.length; + boolean nucleotide = false; + int nucleotideCount = 0; + int peptideCount = 0; - int[] values = new int[255]; + ProfileI[] result = new ProfileI[width]; - char[] seq; - - for (i = start; i < end; i++) + for (int column = start; column < end; column++) { - residueHash = new Hashtable(); - maxCount = 0; - maxResidue = ""; - nongap = 0; - values = new int[255]; + /* + * Apply a heuristic to detect nucleotide data (which can + * be counted in more compact arrays); here we test for + * more than 90% nucleotide; recheck every 10 columns in case + * of misleading data e.g. highly conserved Alanine in peptide! + * Mistakenly guessing nucleotide has a small performance cost, + * as it will result in counting in sparse arrays. + * Mistakenly guessing peptide has a small space cost, + * as it will use a larger than necessary array to hold counts. + */ + if (nucleotideCount > 100 && column % 10 == 0) + { + nucleotide = (9 * peptideCount < nucleotideCount); + } + ResidueCount residueCounts = new ResidueCount(nucleotide); - for (j = 0; j < jSize; j++) + for (int row = 0; row < seqCount; row++) { - if (sequences[j] == null) + if (sequences[row] == null) { - System.err - .println("WARNING: Consensus skipping null sequence - possible race condition."); + System.err.println( + "WARNING: Consensus skipping null sequence - possible race condition."); continue; } - seq = sequences[j].getSequence(); - if (seq.length > i) + if (sequences[row].getLength() > column) { - c = seq[i]; - - if (c == '.' || c == ' ') + char c = sequences[row].getCharAt(column); + residueCounts.add(c); + if (Comparison.isNucleotide(c)) { - c = '-'; + nucleotideCount++; } - - if (c == '-') - { - values['-']++; - continue; - } - else if ('a' <= c && c <= 'z') + else if (!Comparison.isGap(c)) { - c += TO_UPPER_CASE; + peptideCount++; } - - nongap++; - values[c]++; - } else { - values['-']++; + /* + * count a gap if the sequence doesn't reach this column + */ + residueCounts.addGap(); } } - if (jSize == 1) - { - maxResidue = String.valueOf(c); - maxCount = 1; - } - else - { - for (v = 'A'; v <= 'Z'; v++) - { - // TODO why ignore values[v] == 1? - if (values[v] < 1 /* 2 */|| values[v] < maxCount) - { - continue; - } - if (values[v] > maxCount) - { - maxResidue = CHARS[v - 'A']; - } - else if (values[v] == maxCount) - { - maxResidue += CHARS[v - 'A']; - } - maxCount = values[v]; - } - } - if (maxResidue.length() == 0) - { - maxResidue = "-"; - } - if (profile) + int maxCount = residueCounts.getModalCount(); + String maxResidue = residueCounts.getResiduesForCount(maxCount); + int gapCount = residueCounts.getGapCount(); + ProfileI profile = new Profile(seqCount, gapCount, maxCount, + maxResidue); + + if (saveFullProfile) { - // TODO use a 1-dimensional array with jSize, nongap in [0] and [1] - residueHash.put(PROFILE, new int[][] - { values, new int[] - { jSize, nongap } }); + profile.setCounts(residueCounts); } - residueHash.put(MAXCOUNT, new Integer(maxCount)); - residueHash.put(MAXRESIDUE, maxResidue); - percentage = ((float) maxCount * 100) / jSize; - residueHash.put(PID_GAPS, new Float(percentage)); + result[column] = profile; + } + return new Profiles(result); + // long elapsed = System.currentTimeMillis() - now; + // System.out.println(elapsed); + } - if (nongap > 0) + /** + * Returns the full set of profiles for a hidden Markov model. The underlying + * data is the raw probabilities of a residue being emitted at each node, + * however the profiles returned by this function contain the percentage + * chance of a residue emission. + * + * @param hmm + * @param width + * The width of the Profile array (Profiles) to be returned. + * @param start + * The alignment column on which the first profile is based. + * @param end + * The alignment column on which the last profile is based. + * @param removeBelowBackground + * if true, symbols with a match emission probability less than + * background frequency are ignored + * @return + */ + public static ProfilesI calculateHMMProfiles(final HiddenMarkovModel hmm, + int width, int start, int end, boolean removeBelowBackground, + boolean infoLetterHeight) + { + ProfileI[] result = new ProfileI[width]; + char[] symbols = hmm.getSymbols().toCharArray(); + int symbolCount = symbols.length; + for (int column = start; column < end; column++) + { + ResidueCount counts = new ResidueCount(); + for (char symbol : symbols) { - // calculate for non-gapped too - percentage = ((float) maxCount * 100) / nongap; + int value = getAnalogueCount(hmm, column, symbol, + removeBelowBackground, infoLetterHeight); + counts.put(symbol, value); } - residueHash.put(PID_NOGAPS, new Float(percentage)); - - result[i] = residueHash; + int maxCount = counts.getModalCount(); + String maxResidue = counts.getResiduesForCount(maxCount); + int gapCount = counts.getGapCount(); + ProfileI profile = new Profile(symbolCount, gapCount, maxCount, + maxResidue); + profile.setCounts(counts); + + result[column] = profile; } + return new Profiles(result); } /** - * Compute all or part of the annotation row from the given consensus - * hashtable + * Make an estimate of the profile size we are going to compute i.e. how many + * different characters may be present in it. Overestimating has a cost of + * using more memory than necessary. Underestimating has a cost of needing to + * extend the SparseIntArray holding the profile counts. * - * @param consensus - * - pre-allocated annotation row - * @param hconsensus - * @param iStart - * @param width - * @param ignoreGapsInConsensusCalculation - * @param includeAllConsSymbols - * @param nseq + * @param profileSizes + * counts of sizes of profiles so far encountered + * @return */ - public static void completeConsensus(AlignmentAnnotation consensus, - Hashtable[] hconsensus, int iStart, int width, - boolean ignoreGapsInConsensusCalculation, - boolean includeAllConsSymbols, long nseq) + static int estimateProfileSize(SparseIntArray profileSizes) { - completeConsensus(consensus, hconsensus, iStart, width, - ignoreGapsInConsensusCalculation, includeAllConsSymbols, null, - nseq); + if (profileSizes.size() == 0) + { + return 4; + } + + /* + * could do a statistical heuristic here e.g. 75%ile + * for now just return the largest value + */ + return profileSizes.keyAt(profileSizes.size() - 1); } /** * Derive the consensus annotations to be added to the alignment for display. * This does not recompute the raw data, but may be called on a change in - * display options, such as 'show logo', which may in turn result in a change - * in the derived values. + * display options, such as 'ignore gaps', which may in turn result in a + * change in the derived values. * * @param consensus * the annotation row to add annotations to - * @param hconsensus + * @param profiles * the source consensus data - * @param iStart - * start column - * @param width - * end column - * @param ignoreGapsInConsensusCalculation - * if true, use the consensus calculated ignoring gaps - * @param includeAllConsSymbols + * @param startCol + * start column (inclusive) + * @param endCol + * end column (exclusive) + * @param ignoreGaps + * if true, normalise residue percentages ignoring gaps + * @param showSequenceLogo * if true include all consensus symbols, else just show modal * residue - * @param alphabet * @param nseq * number of sequences */ public static void completeConsensus(AlignmentAnnotation consensus, - Hashtable[] hconsensus, int iStart, int width, - boolean ignoreGapsInConsensusCalculation, - boolean includeAllConsSymbols, char[] alphabet, long nseq) + ProfilesI profiles, int startCol, int endCol, boolean ignoreGaps, + boolean showSequenceLogo, long nseq) { + // long now = System.currentTimeMillis(); if (consensus == null || consensus.annotations == null - || consensus.annotations.length < width) + || consensus.annotations.length < endCol) { - // called with a bad alignment annotation row - wait for it to be - // initialised properly + /* + * called with a bad alignment annotation row + * wait for it to be initialised properly + */ return; } - final Format fmt = getPercentageFormat(nseq); - - for (int i = iStart; i < width; i++) + for (int i = startCol; i < endCol; i++) { - Hashtable hci; - if (i >= hconsensus.length || ((hci = hconsensus[i]) == null)) + ProfileI profile = profiles.get(i); + if (profile == null) { - // happens if sequences calculated over were shorter than alignment - // width + /* + * happens if sequences calculated over were + * shorter than alignment width + */ consensus.annotations[i] = null; - continue; + return; } - Float fv = (Float) hci - .get(ignoreGapsInConsensusCalculation ? PID_NOGAPS : PID_GAPS); - if (fv == null) + + final int dp = getPercentageDp(nseq); + + float value = profile.getPercentageIdentity(ignoreGaps); + + String description = getTooltip(profile, value, showSequenceLogo, + ignoreGaps, dp); + + String modalResidue = profile.getModalResidue(); + if ("".equals(modalResidue)) { - consensus.annotations[i] = null; - // data has changed below us .. give up and - continue; + modalResidue = "-"; } - float value = fv.floatValue(); - String maxRes = hci.get(AAFrequency.MAXRESIDUE).toString(); - StringBuilder mouseOver = new StringBuilder(64); - if (maxRes.length() > 1) + else if (modalResidue.length() > 1) { - mouseOver.append("[").append(maxRes).append("] "); - maxRes = "+"; + modalResidue = "+"; } - else + consensus.annotations[i] = new Annotation(modalResidue, description, + ' ', value); + } + // long elapsed = System.currentTimeMillis() - now; + // System.out.println(-elapsed); + } + + /** + * Derive the information annotations to be added to the alignment for + * display. This does not recompute the raw data, but may be called on a + * change in display options, such as 'ignore below background frequency', + * which may in turn result in a change in the derived values. + * + * @param information + * the annotation row to add annotations to + * @param profiles + * the source information data + * @param startCol + * start column (inclusive) + * @param endCol + * end column (exclusive) + * @param ignoreGaps + * if true, normalise residue percentages + * @param showSequenceLogo + * if true include all information symbols, else just show modal + * residue + */ + public static float completeInformation(AlignmentAnnotation information, + ProfilesI profiles, int startCol, int endCol) + { + // long now = System.currentTimeMillis(); + if (information == null || information.annotations == null) + { + /* + * called with a bad alignment annotation row + * wait for it to be initialised properly + */ + return 0; + } + + float max = 0f; + SequenceI hmmSeq = information.sequenceRef; + + int seqLength = hmmSeq.getLength(); + if (information.annotations.length < seqLength) + { + return 0; + } + + HiddenMarkovModel hmm = hmmSeq.getHMM(); + + for (int column = startCol; column < endCol; column++) + { + if (column >= seqLength) { - mouseOver.append(hci.get(AAFrequency.MAXRESIDUE) + " "); + // hmm consensus sequence is shorter than the alignment + break; } - int[][] profile = (int[][]) hci.get(AAFrequency.PROFILE); - if (profile != null && includeAllConsSymbols) + + float value = hmm.getInformationContent(column); + boolean isNaN = Float.isNaN(value); + if (!isNaN) { - int sequenceCount = profile[1][0]; - int nonGappedCount = profile[1][1]; - int normalisedBy = ignoreGapsInConsensusCalculation ? nonGappedCount - : sequenceCount; - mouseOver.setLength(0); - if (alphabet != null) - { - for (int c = 0; c < alphabet.length; c++) - { - float tval = profile[0][alphabet[c]] * 100f / normalisedBy; - mouseOver - .append(((c == 0) ? "" : "; ")) - .append(alphabet[c]) - .append(" ") - .append(((fmt != null) ? fmt.form(tval) : ((int) tval))) - .append("%"); - } - } - else - { - // TODO do this sort once only in calculate()? - // char[][] ca = new char[profile[0].length][]; - char[] ca = new char[profile[0].length]; - float[] vl = new float[profile[0].length]; - for (int c = 0; c < ca.length; c++) - { - ca[c] = (char) c; - // ca[c] = new char[] - // { (char) c }; - vl[c] = profile[0][c]; - } - QuickSort.sort(vl, ca); - for (int p = 0, c = ca.length - 1; profile[0][ca[c]] > 0; c--) - { - final char residue = ca[c]; - if (residue != '-') - { - float tval = profile[0][residue] * 100f / normalisedBy; - mouseOver - .append((((p == 0) ? "" : "; "))) - .append(residue) - .append(" ") - .append(((fmt != null) ? fmt.form(tval) - : ((int) tval))).append("%"); - p++; - } - } - } + max = Math.max(max, value); } - else + + String description = isNaN ? null + : String.format("%.4f bits", value); + information.annotations[column] = new Annotation( + Character.toString( + Character.toUpperCase(hmmSeq.getCharAt(column))), + description, ' ', value); + } + + information.graphMax = max; + return max; + } + + /** + * Derive the occupancy count annotation + * + * @param occupancy + * the annotation row to add annotations to + * @param profiles + * the source consensus data + * @param startCol + * start column (inclusive) + * @param endCol + * end column (exclusive) + */ + public static void completeGapAnnot(AlignmentAnnotation occupancy, + ProfilesI profiles, int startCol, int endCol, long nseq) + { + if (occupancy == null || occupancy.annotations == null + || occupancy.annotations.length < endCol) + { + /* + * called with a bad alignment annotation row + * wait for it to be initialised properly + */ + return; + } + // always set ranges again + occupancy.graphMax = nseq; + occupancy.graphMin = 0; + double scale = 0.8 / nseq; + for (int i = startCol; i < endCol; i++) + { + ProfileI profile = profiles.get(i); + if (profile == null) { - mouseOver.append( - (((fmt != null) ? fmt.form(value) : ((int) value)))) - .append("%"); + /* + * happens if sequences calculated over were + * shorter than alignment width + */ + occupancy.annotations[i] = null; + return; } - consensus.annotations[i] = new Annotation(maxRes, - mouseOver.toString(), ' ', - value); + + final int gapped = profile.getNonGapped(); + + String description = "" + gapped; + + occupancy.annotations[i] = new Annotation("", description, '\0', + gapped, + jalview.util.ColorUtils.bleachColour(Color.DARK_GRAY, + (float) scale * gapped)); } } /** - * Returns a Format designed to show all significant figures for profile - * percentages. For less than 100 sequences, returns null (the integer - * percentage value will be displayed). For 100-999 sequences, returns "%3.1f" + * Returns a tooltip showing either + *
    + *
  • the full profile (percentages of all residues present), if + * showSequenceLogo is true, or
    • + *
  • just the modal (most common) residue(s), if showSequenceLogo is + * false
    • + *
+ * Percentages are as a fraction of all sequence, or only ungapped sequences + * if ignoreGaps is true. * - * @param nseq + * @param profile + * @param pid + * @param showSequenceLogo + * @param ignoreGaps + * @param dp + * the number of decimal places to format percentages to * @return */ - protected static Format getPercentageFormat(long nseq) + static String getTooltip(ProfileI profile, float pid, + boolean showSequenceLogo, boolean ignoreGaps, int dp) { - int scale = 0; - while (nseq >= 10) + ResidueCount counts = profile.getCounts(); + + String description = null; + if (counts != null && showSequenceLogo) { - scale++; - nseq /= 10; + int normaliseBy = ignoreGaps ? profile.getNonGapped() + : profile.getHeight(); + description = counts.getTooltip(normaliseBy, dp); + } + else + { + StringBuilder sb = new StringBuilder(64); + String maxRes = profile.getModalResidue(); + if (maxRes.length() > 1) + { + sb.append("[").append(maxRes).append("]"); + } + else + { + sb.append(maxRes); + } + if (maxRes.length() > 0) + { + sb.append(" "); + Format.appendPercentage(sb, pid, dp); + sb.append("%"); + } + description = sb.toString(); } - return scale <= 1 ? null : new Format("%3." + (scale - 1) + "f"); + return description; } /** @@ -402,46 +509,45 @@ public class AAFrequency * in descending order of percentage value * * - * @param hconsensus - * the data table from which to extract and sort values + * @param profile + * the data object from which to extract and sort values * @param ignoreGaps * if true, only non-gapped values are included in percentage * calculations * @return */ - public static int[] extractProfile(Hashtable hconsensus, - boolean ignoreGaps) + public static int[] extractProfile(ProfileI profile, boolean ignoreGaps) { int[] rtnval = new int[64]; - int[][] profile = (int[][]) hconsensus.get(AAFrequency.PROFILE); - if (profile == null) + ResidueCount counts = profile.getCounts(); + if (counts == null) { return null; } - char[] ca = new char[profile[0].length]; - float[] vl = new float[profile[0].length]; - for (int c = 0; c < ca.length; c++) - { - ca[c] = (char) c; - vl[c] = profile[0][c]; - } - QuickSort.sort(vl, ca); + + SymbolCounts symbolCounts = counts.getSymbolCounts(); + char[] symbols = symbolCounts.symbols; + int[] values = symbolCounts.values; + QuickSort.sort(values, symbols); int nextArrayPos = 2; int totalPercentage = 0; - int distinctValuesCount = 0; - final int divisor = profile[1][ignoreGaps ? 1 : 0]; - for (int c = ca.length - 1; profile[0][ca[c]] > 0; c--) + final int divisor = ignoreGaps ? profile.getNonGapped() + : profile.getHeight(); + + /* + * traverse the arrays in reverse order (highest counts first) + */ + for (int i = symbols.length - 1; i >= 0; i--) { - if (ca[c] != '-') - { - rtnval[nextArrayPos++] = ca[c]; - final int percentage = (int) (profile[0][ca[c]] * 100f / divisor); - rtnval[nextArrayPos++] = percentage; - totalPercentage += percentage; - distinctValuesCount++; - } + int theChar = symbols[i]; + int charCount = values[i]; + + rtnval[nextArrayPos++] = theChar; + final int percentage = (charCount * 100) / divisor; + rtnval[nextArrayPos++] = percentage; + totalPercentage += percentage; } - rtnval[0] = distinctValuesCount; + rtnval[0] = symbols.length; rtnval[1] = totalPercentage; int[] result = new int[rtnval.length + 1]; result[0] = AlignmentAnnotation.SEQUENCE_PROFILE; @@ -450,6 +556,7 @@ public class AAFrequency return result; } + /** * Extract a sorted extract of cDNA codon profile data. The returned array * contains @@ -462,7 +569,8 @@ public class AAFrequency * @param hashtable * @return */ - public static int[] extractCdnaProfile(Hashtable hashtable, boolean ignoreGaps) + public static int[] extractCdnaProfile(Hashtable hashtable, + boolean ignoreGaps) { // this holds #seqs, #ungapped, and then codon count, indexed by encoded // codon triplet @@ -507,7 +615,7 @@ public class AAFrequency result[1] = distinctValuesCount; return Arrays.copyOfRange(result, 0, j); } - + /** * Compute a consensus for the cDNA coding for a protein alignment. * @@ -521,7 +629,7 @@ public class AAFrequency Hashtable[] hconsensus) { final char gapCharacter = alignment.getGapCharacter(); - Set mappings = alignment.getCodonFrames(); + List mappings = alignment.getCodonFrames(); if (mappings == null || mappings.isEmpty()) { return; @@ -531,7 +639,7 @@ public class AAFrequency for (int col = 0; col < cols; col++) { // todo would prefer a Java bean for consensus data - Hashtable columnHash = new Hashtable(); + Hashtable columnHash = new Hashtable<>(); // #seqs, #ungapped seqs, counts indexed by (codon encoded + 1) int[] codonCounts = new int[66]; codonCounts[0] = alignment.getSequences().size(); @@ -542,12 +650,16 @@ public class AAFrequency { continue; } - char[] codon = MappingUtils.findCodonFor(seq, col, mappings); - int codonEncoded = CodingUtils.encodeCodon(codon); - if (codonEncoded >= 0) + List codons = MappingUtils.findCodonsFor(seq, col, + mappings); + for (char[] codon : codons) { - codonCounts[codonEncoded + 2]++; - ungappedCount++; + int codonEncoded = CodingUtils.encodeCodon(codon); + if (codonEncoded >= 0) + { + codonCounts[codonEncoded + 2]++; + ungappedCount++; + } } } codonCounts[1] = ungappedCount; @@ -619,11 +731,14 @@ public class AAFrequency int modalCodonEncoded = codons[codons.length - 1]; int modalCodonCount = sortedCodonCounts[codons.length - 1]; - String modalCodon = String.valueOf(CodingUtils - .decodeCodon(modalCodonEncoded)); - if (sortedCodonCounts.length > 1 - && sortedCodonCounts[codons.length - 2] == modalCodonEncoded) + String modalCodon = String + .valueOf(CodingUtils.decodeCodon(modalCodonEncoded)); + if (sortedCodonCounts.length > 1 && sortedCodonCounts[codons.length + - 2] == sortedCodonCounts[codons.length - 1]) { + /* + * two or more codons share the modal count + */ modalCodon = "+"; } float pid = sortedCodonCounts[sortedCodonCounts.length - 1] * 100 @@ -642,7 +757,7 @@ public class AAFrequency StringBuilder samePercent = new StringBuilder(); String percent = null; String lastPercent = null; - Format fmt = getPercentageFormat(nseqs); + int percentDecPl = getPercentageDp(nseqs); for (int j = codons.length - 1; j >= 0; j--) { @@ -664,8 +779,9 @@ public class AAFrequency final int pct = codonCount * 100 / totalCount; String codon = String .valueOf(CodingUtils.decodeCodon(codonEncoded)); - percent = fmt == null ? Integer.toString(pct) : fmt - .form(pct); + StringBuilder sb = new StringBuilder(); + Format.appendPercentage(sb, pct, percentDecPl); + percent = sb.toString(); if (showProfileLogo || codonCount == modalCodonCount) { if (percent.equals(lastPercent) && j > 0) @@ -677,8 +793,7 @@ public class AAFrequency { if (samePercent.length() > 0) { - mouseOver.append(samePercent).append(": ") - .append(lastPercent) + mouseOver.append(samePercent).append(": ").append(lastPercent) .append("% "); } samePercent.setLength(0); @@ -692,4 +807,137 @@ public class AAFrequency mouseOver.toString(), ' ', pid); } } + + /** + * Returns the number of decimal places to show for profile percentages. For + * less than 100 sequences, returns zero (the integer percentage value will be + * displayed). For 100-999 sequences, returns 1, for 1000-9999 returns 2, etc. + * + * @param nseq + * @return + */ + protected static int getPercentageDp(long nseq) + { + int scale = 0; + while (nseq >= 100) + { + scale++; + nseq /= 10; + } + return scale; + } + + /** + * Returns the sorted HMM profile for the given column of the alignment. The + * returned array contains + * + * 
+   *    [profileType=0, numberOfValues, 100, charValue1, percentage1, charValue2, percentage2, ...]
+   * in descending order of percentage value
+   *
+ * + * @param hmm + * @param column + * @param removeBelowBackground + * if true, ignores residues with probability less than their + * background frequency + * @param infoHeight + * if true, uses the log ratio 'information' measure to scale the + * value + * @return + */ + public static int[] extractHMMProfile(HiddenMarkovModel hmm, int column, + boolean removeBelowBackground, boolean infoHeight) + { + if (hmm == null) + { + return null; + } + String alphabet = hmm.getSymbols(); + int size = alphabet.length(); + char symbols[] = new char[size]; + int values[] = new int[size]; + int totalCount = 0; + + for (int i = 0; i < size; i++) + { + char symbol = alphabet.charAt(i); + symbols[i] = symbol; + int value = getAnalogueCount(hmm, column, symbol, + removeBelowBackground, infoHeight); + values[i] = value; + totalCount += value; + } + + /* + * sort symbols by increasing emission probability + */ + QuickSort.sort(values, symbols); + + int[] profile = new int[3 + size * 2]; + + profile[0] = AlignmentAnnotation.SEQUENCE_PROFILE; + profile[1] = size; + profile[2] = 100; + + /* + * order symbol/count profile by decreasing emission probability + */ + if (totalCount != 0) + { + int arrayPos = 3; + for (int k = size - 1; k >= 0; k--) + { + Float percentage; + int value = values[k]; + if (removeBelowBackground) + { + percentage = ((float) value) / totalCount * 100f; + } + else + { + percentage = value / 100f; + } + int intPercent = Math.round(percentage); + profile[arrayPos] = symbols[k]; + profile[arrayPos + 1] = intPercent; + arrayPos += 2; + } + } + return profile; + } + + /** + * Converts the emission probability of a residue at a column in the alignment + * to a 'count', suitable for rendering as an annotation value + * + * @param hmm + * @param column + * @param symbol + * @param removeBelowBackground + * if true, returns 0 for any symbol with a match emission + * probability less than the background frequency + * @infoHeight if true, uses the log ratio 'information content' to scale the + * value + * @return + */ + static int getAnalogueCount(HiddenMarkovModel hmm, int column, + char symbol, boolean removeBelowBackground, boolean infoHeight) + { + double value = hmm.getMatchEmissionProbability(column, symbol); + double freq = ResidueProperties.backgroundFrequencies + .get(hmm.getAlphabetType()).get(symbol); + if (value < freq && removeBelowBackground) + { + return 0; + } + + if (infoHeight) + { + value = value * (Math.log(value / freq) / LOG2); + } + + value = value * 10000d; + return Math.round((float) value); + } }