X-Git-Url: http://source.jalview.org/gitweb/?a=blobdiff_plain;f=src%2Fjalview%2Fanalysis%2FAAFrequency.java;h=892b1b1ff8cb4efc79cdb7f3bac0ba3a6542bb7e;hb=d043ce47fc710d3eb2629ba926a8a7417bd67d8c;hp=5ecd64426cd87203db003444d9d7f556ba35e6d8;hpb=a5ddd79bb9c1fcc97ffbe01e9e5ed8caa5d7b435;p=jalview.git diff --git a/src/jalview/analysis/AAFrequency.java b/src/jalview/analysis/AAFrequency.java index 5ecd644..892b1b1 100755 --- a/src/jalview/analysis/AAFrequency.java +++ b/src/jalview/analysis/AAFrequency.java @@ -20,18 +20,26 @@ */ package jalview.analysis; -import jalview.analysis.ResidueCount.SymbolCounts; 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; @@ -42,43 +50,20 @@ import java.util.List; * 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 { - 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 Profile[] 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 Profile[] calculate(List sequences, + public static final ProfilesI calculate(List sequences, int start, int end, boolean profile) { SequenceI[] seqs = new SequenceI[sequences.size()]; @@ -88,20 +73,19 @@ 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; } } - Profile[] reply = new Profile[width]; - if (end >= width) { end = width; } - calculate(seqs, start, end, reply, profile); + ProfilesI reply = calculate(seqs, width, start, end, profile); return reply; } } @@ -110,15 +94,17 @@ public class AAFrequency * 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 - * @param result - * array in which to store profile per column + * end column (exclusive) * @param saveFullProfile * if true, store all symbol counts */ - public static final void calculate(final SequenceI[] sequences, - int start, int end, Profile[] result, boolean saveFullProfile) + public static final ProfilesI calculate(final SequenceI[] sequences, + int width, int start, int end, boolean saveFullProfile) { // long now = System.currentTimeMillis(); int seqCount = sequences.length; @@ -126,6 +112,8 @@ public class AAFrequency int nucleotideCount = 0; int peptideCount = 0; + ProfileI[] result = new ProfileI[width]; + for (int column = start; column < end; column++) { /* @@ -148,14 +136,13 @@ public class AAFrequency { 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; } - char[] seq = sequences[row].getSequence(); - if (seq.length > column) + if (sequences[row].getLength() > column) { - char c = seq[column]; + char c = sequences[row].getCharAt(column); residueCounts.add(c); if (Comparison.isNucleotide(c)) { @@ -169,8 +156,7 @@ public class AAFrequency else { /* - * here we count a gap if the sequence doesn't - * reach this column (is that correct?) + * count a gap if the sequence doesn't reach this column */ residueCounts.addGap(); } @@ -179,7 +165,7 @@ public class AAFrequency int maxCount = residueCounts.getModalCount(); String maxResidue = residueCounts.getResiduesForCount(maxCount); int gapCount = residueCounts.getGapCount(); - Profile profile = new Profile(seqCount, gapCount, maxCount, + ProfileI profile = new Profile(seqCount, gapCount, maxCount, maxResidue); if (saveFullProfile) @@ -189,11 +175,58 @@ public class AAFrequency result[column] = profile; } + return new Profiles(result); // long elapsed = System.currentTimeMillis() - now; // System.out.println(elapsed); } /** + * 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) + { + int value = getAnalogueCount(hmm, column, symbol, + removeBelowBackground, infoLetterHeight); + counts.put(symbol, value); + } + 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); + } + + /** * 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 @@ -220,17 +253,17 @@ public class AAFrequency /** * 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 profiles * the source consensus data - * @param iStart - * start column - * @param width - * end column + * @param startCol + * start column (inclusive) + * @param endCol + * end column (exclusive) * @param ignoreGaps * if true, normalise residue percentages ignoring gaps * @param showSequenceLogo @@ -240,12 +273,12 @@ public class AAFrequency * number of sequences */ public static void completeConsensus(AlignmentAnnotation consensus, - Profile[] profiles, int iStart, int width, boolean ignoreGaps, + 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 @@ -254,39 +287,172 @@ public class AAFrequency return; } - final int dp = getPercentageDp(nseq); - - for (int i = iStart; i < width; i++) + for (int i = startCol; i < endCol; i++) { - Profile profile; - if (i >= profiles.length || ((profile = profiles[i]) == null)) + ProfileI profile = profiles.get(i); + if (profile == null) { /* * happens if sequences calculated over were * shorter than alignment width */ consensus.annotations[i] = null; - continue; + return; } + final int dp = getPercentageDp(nseq); + float value = profile.getPercentageIdentity(ignoreGaps); String description = getTooltip(profile, value, showSequenceLogo, ignoreGaps, dp); - consensus.annotations[i] = new Annotation(profile.getModalResidue(), - description, ' ', value); + String modalResidue = profile.getModalResidue(); + if ("".equals(modalResidue)) + { + modalResidue = "-"; + } + else if (modalResidue.length() > 1) + { + modalResidue = "+"; + } + 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) + { + // hmm consensus sequence is shorter than the alignment + break; + } + + float value = hmm.getInformationContent(column); + boolean isNaN = Float.isNaN(value); + if (!isNaN) + { + max = Math.max(max, value); + } + + 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) + { + /* + * happens if sequences calculated over were + * shorter than alignment width + */ + occupancy.annotations[i] = null; + return; + } + + 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 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
    • + *
  • 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. @@ -299,7 +465,7 @@ public class AAFrequency * the number of decimal places to format percentages to * @return */ - static String getTooltip(Profile profile, float pid, + static String getTooltip(ProfileI profile, float pid, boolean showSequenceLogo, boolean ignoreGaps, int dp) { ResidueCount counts = profile.getCounts(); @@ -307,8 +473,8 @@ public class AAFrequency String description = null; if (counts != null && showSequenceLogo) { - int normaliseBy = ignoreGaps ? profile.getNonGapped() : profile - .getHeight(); + int normaliseBy = ignoreGaps ? profile.getNonGapped() + : profile.getHeight(); description = counts.getTooltip(normaliseBy, dp); } else @@ -317,15 +483,18 @@ public class AAFrequency String maxRes = profile.getModalResidue(); if (maxRes.length() > 1) { - sb.append("[").append(maxRes).append("] "); - maxRes = "+"; + sb.append("[").append(maxRes).append("]"); } else { - sb.append(maxRes).append(" "); + sb.append(maxRes); + } + if (maxRes.length() > 0) + { + sb.append(" "); + Format.appendPercentage(sb, pid, dp); + sb.append("%"); } - Format.appendPercentage(sb, pid, dp); - sb.append("%"); description = sb.toString(); } return description; @@ -336,7 +505,7 @@ public class AAFrequency * contains * * 
-   *    [profileType, numberOfValues, nonGapCount, charValue1, percentage1, charValue2, percentage2, ...]
+   *    [profileType, numberOfValues, totalPercent, charValue1, percentage1, charValue2, percentage2, ...]
    * in descending order of percentage value
    *
* @@ -347,10 +516,8 @@ public class AAFrequency * calculations * @return */ - public static int[] extractProfile(Profile profile, - boolean ignoreGaps) + public static int[] extractProfile(ProfileI profile, boolean ignoreGaps) { - int[] rtnval = new int[64]; ResidueCount counts = profile.getCounts(); if (counts == null) { @@ -361,47 +528,70 @@ public class AAFrequency char[] symbols = symbolCounts.symbols; int[] values = symbolCounts.values; QuickSort.sort(values, symbols); - int nextArrayPos = 2; int totalPercentage = 0; - final int divisor = ignoreGaps ? profile.getNonGapped() : profile - .getHeight(); + final int divisor = ignoreGaps ? profile.getNonGapped() + : profile.getHeight(); /* * traverse the arrays in reverse order (highest counts first) */ + int[] result = new int[3 + 2 * symbols.length]; + int nextArrayPos = 3; + int nonZeroCount = 0; + for (int i = symbols.length - 1; i >= 0; i--) { int theChar = symbols[i]; int charCount = values[i]; - - rtnval[nextArrayPos++] = theChar; final int percentage = (charCount * 100) / divisor; - rtnval[nextArrayPos++] = percentage; + if (percentage == 0) + { + /* + * this count (and any remaining) round down to 0% - discard + */ + break; + } + nonZeroCount++; + result[nextArrayPos++] = theChar; + result[nextArrayPos++] = percentage; totalPercentage += percentage; } - rtnval[0] = symbols.length; - rtnval[1] = totalPercentage; - int[] result = new int[rtnval.length + 1]; + + /* + * truncate array if any zero values were discarded + */ + if (nonZeroCount < symbols.length) + { + int[] tmp = new int[3 + 2 * nonZeroCount]; + System.arraycopy(result, 0, tmp, 0, tmp.length); + result = tmp; + } + + /* + * fill in 'header' values + */ result[0] = AlignmentAnnotation.SEQUENCE_PROFILE; - System.arraycopy(rtnval, 0, result, 1, rtnval.length); + result[1] = nonZeroCount; + result[2] = totalPercentage; return result; } + /** * Extract a sorted extract of cDNA codon profile data. The returned array * contains * * 
-   *    [profileType, numberOfValues, totalCount, charValue1, percentage1, charValue2, percentage2, ...]
+   *    [profileType, numberOfValues, totalPercentage, charValue1, percentage1, charValue2, percentage2, ...]
    * in descending order of percentage value, where the character values encode codon triplets
    *
* * @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 @@ -431,9 +621,16 @@ public class AAFrequency { break; // nothing else of interest here } + final int percentage = codonCount * 100 / divisor; + if (percentage == 0) + { + /* + * this (and any remaining) values rounded down to 0 - discard + */ + break; + } distinctValuesCount++; result[j++] = codons[i]; - final int percentage = codonCount * 100 / divisor; result[j++] = percentage; totalPercentage += percentage; } @@ -457,7 +654,7 @@ public class AAFrequency * the consensus data stores to be populated (one per column) */ public static void calculateCdna(AlignmentI alignment, - Hashtable[] hconsensus) + Hashtable[] hconsensus) { final char gapCharacter = alignment.getGapCharacter(); List mappings = alignment.getCodonFrames(); @@ -470,7 +667,7 @@ 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(); @@ -481,8 +678,8 @@ public class AAFrequency { continue; } - List codons = MappingUtils - .findCodonsFor(seq, col, mappings); + List codons = MappingUtils.findCodonsFor(seq, col, + mappings); for (char[] codon : codons) { int codonEncoded = CodingUtils.encodeCodon(codon); @@ -490,6 +687,7 @@ public class AAFrequency { codonCounts[codonEncoded + 2]++; ungappedCount++; + break; } } } @@ -515,7 +713,8 @@ public class AAFrequency */ public static void completeCdnaConsensus( AlignmentAnnotation consensusAnnotation, - Hashtable[] consensusData, boolean showProfileLogo, int nseqs) + Hashtable[] consensusData, + boolean showProfileLogo, int nseqs) { if (consensusAnnotation == null || consensusAnnotation.annotations == null @@ -530,7 +729,7 @@ public class AAFrequency consensusAnnotation.scaleColLabel = true; for (int col = 0; col < consensusData.length; col++) { - Hashtable hci = consensusData[col]; + Hashtable hci = consensusData[col]; if (hci == null) { // gapped protein column? @@ -562,10 +761,10 @@ 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] == sortedCodonCounts[codons.length - 1]) + 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 @@ -624,8 +823,8 @@ public class AAFrequency { if (samePercent.length() > 0) { - mouseOver.append(samePercent).append(": ") - .append(lastPercent).append("% "); + mouseOver.append(samePercent).append(": ").append(lastPercent) + .append("% "); } samePercent.setLength(0); samePercent.append(codon); @@ -657,4 +856,118 @@ public class AAFrequency } 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); + } }