X-Git-Url: http://source.jalview.org/gitweb/?a=blobdiff_plain;ds=inline;f=src%2Fjalview%2Fanalysis%2FAlignmentUtils.java;h=6f0125d97eb5e287aae5f90fc8d9a359375bd1d1;hb=f18077aa57dea5481ae537162871045f8557e2f0;hp=74406dc2d340bfa46a4c4c98597ebd784058afbf;hpb=4924c107e70d763821067bcb3e1586bc14589918;p=jalview.git diff --git a/src/jalview/analysis/AlignmentUtils.java b/src/jalview/analysis/AlignmentUtils.java index 74406dc..6f0125d 100644 --- a/src/jalview/analysis/AlignmentUtils.java +++ b/src/jalview/analysis/AlignmentUtils.java @@ -1,6 +1,6 @@ /* - * Jalview - A Sequence Alignment Editor and Viewer (Version 2.8.2) - * Copyright (C) 2014 The Jalview Authors + * Jalview - A Sequence Alignment Editor and Viewer ($$Version-Rel$$) + * Copyright (C) $$Year-Rel$$ The Jalview Authors * * This file is part of Jalview. * @@ -20,20 +20,11 @@ */ package jalview.analysis; -import jalview.datamodel.AlignedCodon; -import jalview.datamodel.AlignedCodonFrame; -import jalview.datamodel.AlignmentAnnotation; -import jalview.datamodel.AlignmentI; -import jalview.datamodel.Mapping; -import jalview.datamodel.SearchResults; -import jalview.datamodel.Sequence; -import jalview.datamodel.SequenceI; -import jalview.schemes.ResidueProperties; -import jalview.util.MapList; - import java.util.ArrayList; import java.util.Arrays; +import java.util.Collection; import java.util.HashMap; +import java.util.HashSet; import java.util.Iterator; import java.util.LinkedHashMap; import java.util.List; @@ -42,6 +33,18 @@ import java.util.Map.Entry; import java.util.Set; import java.util.TreeMap; +import jalview.datamodel.AlignedCodon; +import jalview.datamodel.AlignedCodonFrame; +import jalview.datamodel.AlignmentAnnotation; +import jalview.datamodel.AlignmentI; +import jalview.datamodel.Mapping; +import jalview.datamodel.SearchResults; +import jalview.datamodel.Sequence; +import jalview.datamodel.SequenceGroup; +import jalview.datamodel.SequenceI; +import jalview.schemes.ResidueProperties; +import jalview.util.MapList; + /** * grab bag of useful alignment manipulation operations Expect these to be * refactored elsewhere at some point. @@ -53,15 +56,6 @@ public class AlignmentUtils { /** - * Represents the 3 possible results of trying to map one alignment to - * another. - */ - public enum MappingResult - { - Mapped, NotMapped, AlreadyMapped - } - - /** * given an existing alignment, create a new alignment including all, or up to * flankSize additional symbols from each sequence's dataset sequence * @@ -214,51 +208,95 @@ public class AlignmentUtils /** * Build mapping of protein to cDNA alignment. Mappings are made between - * sequences which have the same name and compatible lengths. Any new mappings - * are added to the protein alignment. Has a 3-valued result: either Mapped - * (at least one sequence mapping was created), AlreadyMapped (all possible - * sequence mappings already exist), or NotMapped (no possible sequence - * mappings exist). + * sequences where the cDNA translates to the protein sequence. Any new + * mappings are added to the protein alignment. Returns true if any mappings + * either already exist or were added, else false. * * @param proteinAlignment * @param cdnaAlignment * @return */ - public static MappingResult mapProteinToCdna( + public static boolean mapProteinToCdna( final AlignmentI proteinAlignment, final AlignmentI cdnaAlignment) { if (proteinAlignment == null || cdnaAlignment == null) { - return MappingResult.NotMapped; + return false; } - boolean mappingPossible = false; - boolean mappingPerformed = false; + Set mappedDna = new HashSet(); + Set mappedProtein = new HashSet(); - List thisSeqs = proteinAlignment.getSequences(); - /* - * Build a look-up of cDNA sequences by name, for matching purposes. + * First pass - map sequences where cross-references exist. This include + * 1-to-many mappings to support, for example, variant cDNA. */ - Map> cdnaSeqs = cdnaAlignment - .getSequencesByName(); - + boolean mappingPerformed = mapProteinToCdna(proteinAlignment, + cdnaAlignment, mappedDna, mappedProtein, true); + + /* + * Second pass - map sequences where no cross-references exist. This only + * does 1-to-1 mappings and assumes corresponding sequences are in the same + * order in the alignments. + */ + mappingPerformed |= mapProteinToCdna(proteinAlignment, cdnaAlignment, + mappedDna, mappedProtein, false); + return mappingPerformed; + } + + /** + * Make mappings between compatible sequences (where the cDNA translation + * matches the protein). + * + * @param proteinAlignment + * @param cdnaAlignment + * @param mappedDna + * a set of mapped DNA sequences (to add to) + * @param mappedProtein + * a set of mapped Protein sequences (to add to) + * @param xrefsOnly + * if true, only map sequences where xrefs exist + * @return + */ + protected static boolean mapProteinToCdna( + final AlignmentI proteinAlignment, + final AlignmentI cdnaAlignment, Set mappedDna, + Set mappedProtein, boolean xrefsOnly) + { + boolean mappingPerformed = false; + List thisSeqs = proteinAlignment.getSequences(); for (SequenceI aaSeq : thisSeqs) { + boolean proteinMapped = false; AlignedCodonFrame acf = new AlignedCodonFrame(); - List candidates = cdnaSeqs.get(aaSeq.getName()); - if (candidates == null) + + for (SequenceI cdnaSeq : cdnaAlignment.getSequences()) { /* - * No cDNA sequence with matching name, so no mapping possible for this - * protein sequence + * Always try to map if sequences have xref to each other; this supports + * variant cDNA or alternative splicing for a protein sequence. + * + * If no xrefs, try to map progressively, assuming that alignments have + * mappable sequences in corresponding order. These are not + * many-to-many, as that would risk mixing species with similar cDNA + * sequences. */ - continue; - } - mappingPossible = true; - for (SequenceI cdnaSeq : candidates) - { + if (xrefsOnly && !CrossRef.haveCrossRef(aaSeq, cdnaSeq)) + { + continue; + } + + /* + * Don't map non-xrefd sequences more than once each. This heuristic + * allows us to pair up similar sequences in ordered alignments. + */ + if (!xrefsOnly + && (mappedProtein.contains(aaSeq) || mappedDna + .contains(cdnaSeq))) + { + continue; + } if (!mappingExists(proteinAlignment.getCodonFrames(), aaSeq.getDatasetSequence(), cdnaSeq.getDatasetSequence())) { @@ -267,25 +305,18 @@ public class AlignmentUtils { acf.addMap(cdnaSeq, aaSeq, map); mappingPerformed = true; + proteinMapped = true; + mappedDna.add(cdnaSeq); + mappedProtein.add(aaSeq); } } } - proteinAlignment.addCodonFrame(acf); - } - - /* - * If at least one mapping was possible but none was done, then the - * alignments are already as mapped as they can be. - */ - if (mappingPossible && !mappingPerformed) - { - return MappingResult.AlreadyMapped; - } - else - { - return mappingPerformed ? MappingResult.Mapped - : MappingResult.NotMapped; + if (proteinMapped) + { + proteinAlignment.addCodonFrame(acf); + } } + return mappingPerformed; } /** @@ -311,7 +342,8 @@ public class AlignmentUtils /** * Build a mapping (if possible) of a protein to a cDNA sequence. The cDNA * must be three times the length of the protein, possibly after ignoring - * start and/or stop codons. Returns null if no mapping is determined. + * start and/or stop codons, and must translate to the protein. Returns null + * if no mapping is determined. * * @param proteinSeqs * @param cdnaSeq @@ -321,34 +353,41 @@ public class AlignmentUtils SequenceI cdnaSeq) { /* - * Here we handle either dataset sequence set (desktop) or absent (applet) + * Here we handle either dataset sequence set (desktop) or absent (applet). + * Use only the char[] form of the sequence to avoid creating possibly large + * String objects. */ final SequenceI proteinDataset = proteinSeq.getDatasetSequence(); - String aaSeqString = proteinDataset != null ? proteinDataset - .getSequenceAsString() : proteinSeq.getSequenceAsString(); + char[] aaSeqChars = proteinDataset != null ? proteinDataset + .getSequence() : proteinSeq.getSequence(); final SequenceI cdnaDataset = cdnaSeq.getDatasetSequence(); - String cdnaSeqString = cdnaDataset != null ? cdnaDataset - .getSequenceAsString() : cdnaSeq.getSequenceAsString(); - if (aaSeqString == null || cdnaSeqString == null) + char[] cdnaSeqChars = cdnaDataset != null ? cdnaDataset.getSequence() + : cdnaSeq.getSequence(); + if (aaSeqChars == null || cdnaSeqChars == null) { return null; } - final int mappedLength = 3 * aaSeqString.length(); - int cdnaLength = cdnaSeqString.length(); + /* + * cdnaStart/End, proteinStartEnd are base 1 (for dataset sequence mapping) + */ + final int mappedLength = 3 * aaSeqChars.length; + int cdnaLength = cdnaSeqChars.length; int cdnaStart = 1; int cdnaEnd = cdnaLength; final int proteinStart = 1; - final int proteinEnd = aaSeqString.length(); + final int proteinEnd = aaSeqChars.length; /* * If lengths don't match, try ignoring stop codon. */ - if (cdnaLength != mappedLength) + if (cdnaLength != mappedLength && cdnaLength > 2) { - for (Object stop : ResidueProperties.STOP) + String lastCodon = String.valueOf(cdnaSeqChars, cdnaLength - 3, 3) + .toUpperCase(); + for (String stop : ResidueProperties.STOP) { - if (cdnaSeqString.toUpperCase().endsWith((String) stop)) + if (lastCodon.equals(stop)) { cdnaEnd -= 3; cdnaLength -= 3; @@ -361,24 +400,68 @@ public class AlignmentUtils * If lengths still don't match, try ignoring start codon. */ if (cdnaLength != mappedLength - && cdnaSeqString.toUpperCase().startsWith( + && cdnaLength > 2 + && String.valueOf(cdnaSeqChars, 0, 3).toUpperCase() + .equals( ResidueProperties.START)) { cdnaStart += 3; cdnaLength -= 3; } - if (cdnaLength == mappedLength) + if (cdnaLength != mappedLength) { - MapList map = new MapList(new int[] - { cdnaStart, cdnaEnd }, new int[] - { proteinStart, proteinEnd }, 3, 1); - return map; + return null; } - else + if (!translatesAs(cdnaSeqChars, cdnaStart - 1, aaSeqChars)) { return null; } + MapList map = new MapList(new int[] + { cdnaStart, cdnaEnd }, new int[] + { proteinStart, proteinEnd }, 3, 1); + return map; + } + + /** + * Test whether the given cdna sequence, starting at the given offset, + * translates to the given amino acid sequence, using the standard translation + * table. Designed to fail fast i.e. as soon as a mismatch position is found. + * + * @param cdnaSeqChars + * @param cdnaStart + * @param aaSeqChars + * @return + */ + protected static boolean translatesAs(char[] cdnaSeqChars, int cdnaStart, + char[] aaSeqChars) + { + int aaResidue = 0; + for (int i = cdnaStart; i < cdnaSeqChars.length - 2 + && aaResidue < aaSeqChars.length; i += 3, aaResidue++) + { + String codon = String.valueOf(cdnaSeqChars, i, 3); + final String translated = ResidueProperties.codonTranslate( + codon); + /* + * ? allow X in protein to match untranslatable in dna ? + */ + final char aaRes = aaSeqChars[aaResidue]; + if ((translated == null || "STOP".equals(translated)) && aaRes == 'X') + { + continue; + } + if (translated == null + || !(aaRes == translated.charAt(0))) + { + // debug + // System.out.println(("Mismatch at " + i + "/" + aaResidue + ": " + // + codon + "(" + translated + ") != " + aaRes)); + return false; + } + } + // fail if we didn't match all of the aa sequence + return (aaResidue == aaSeqChars.length); } /** @@ -405,7 +488,7 @@ public class AlignmentUtils // TODO there may be one AlignedCodonFrame per dataset sequence, or one with // all mappings. Would it help to constrain this? List mappings = al.getCodonFrame(seq); - if (mappings == null) + if (mappings == null || mappings.isEmpty()) { return false; } @@ -902,4 +985,231 @@ public class AlignmentUtils seqProduct.put(protein, codon.product); } } + + /** + * Returns true if a cDNA/Protein mapping either exists, or could be made, + * between at least one pair of sequences in the two alignments. Currently, + * the logic is: + *
    + *
  • One alignment must be nucleotide, and the other protein
    • + *
  • At least one pair of sequences must be already mapped, or mappable
    • + *
  • Mappable means the nucleotide translation matches the protein sequence
    • + *
  • The translation may ignore start and stop codons if present in the + * nucleotide
    • + *
+ * + * @param al1 + * @param al2 + * @return + */ + public static boolean isMappable(AlignmentI al1, AlignmentI al2) + { + /* + * Require one nucleotide and one protein + */ + if (al1.isNucleotide() == al2.isNucleotide()) + { + return false; + } + AlignmentI dna = al1.isNucleotide() ? al1 : al2; + AlignmentI protein = dna == al1 ? al2 : al1; + Set mappings = protein.getCodonFrames(); + for (SequenceI dnaSeq : dna.getSequences()) + { + for (SequenceI proteinSeq : protein.getSequences()) + { + if (isMappable(dnaSeq, proteinSeq, mappings)) + { + return true; + } + } + } + return false; + } + + /** + * Returns true if the dna sequence is mapped, or could be mapped, to the + * protein sequence. + * + * @param dnaSeq + * @param proteinSeq + * @param mappings + * @return + */ + public static boolean isMappable(SequenceI dnaSeq, SequenceI proteinSeq, + Set mappings) + { + SequenceI dnaDs = dnaSeq.getDatasetSequence() == null ? dnaSeq : dnaSeq.getDatasetSequence(); + SequenceI proteinDs = proteinSeq.getDatasetSequence() == null ? proteinSeq + : proteinSeq.getDatasetSequence(); + + /* + * Already mapped? + */ + for (AlignedCodonFrame mapping : mappings) { + if ( proteinDs == mapping.getAaForDnaSeq(dnaDs)) { + return true; + } + } + + /* + * Just try to make a mapping (it is not yet stored), test whether + * successful. + */ + return mapProteinToCdna(proteinDs, dnaDs) != null; + } + + /** + * Finds any reference annotations associated with the sequences in + * sequenceScope, that are not already added to the alignment, and adds them + * to the 'candidates' map. Also populates a lookup table of annotation + * labels, keyed by calcId, for use in constructing tooltips or the like. + * + * @param sequenceScope + * the sequences to scan for reference annotations + * @param labelForCalcId + * (optional) map to populate with label for calcId + * @param candidates + * map to populate with annotations for sequence + * @param al + * the alignment to check for presence of annotations + */ + public static void findAddableReferenceAnnotations( + List sequenceScope, Map labelForCalcId, + final Map> candidates, + AlignmentI al) + { + if (sequenceScope == null) + { + return; + } + + /* + * For each sequence in scope, make a list of any annotations on the + * underlying dataset sequence which are not already on the alignment. + * + * Add to a map of { alignmentSequence, } + */ + for (SequenceI seq : sequenceScope) + { + SequenceI dataset = seq.getDatasetSequence(); + if (dataset == null) + { + continue; + } + AlignmentAnnotation[] datasetAnnotations = dataset.getAnnotation(); + if (datasetAnnotations == null) + { + continue; + } + final List result = new ArrayList(); + for (AlignmentAnnotation dsann : datasetAnnotations) + { + /* + * Find matching annotations on the alignment. If none is found, then + * add this annotation to the list of 'addable' annotations for this + * sequence. + */ + final Iterable matchedAlignmentAnnotations = al + .findAnnotations(seq, dsann.getCalcId(), + dsann.label); + if (!matchedAlignmentAnnotations.iterator().hasNext()) + { + result.add(dsann); + if (labelForCalcId != null) + { + labelForCalcId.put(dsann.getCalcId(), dsann.label); + } + } + } + /* + * Save any addable annotations for this sequence + */ + if (!result.isEmpty()) + { + candidates.put(seq, result); + } + } + } + + /** + * Adds annotations to the top of the alignment annotations, in the same order + * as their related sequences. + * + * @param annotations + * the annotations to add + * @param alignment + * the alignment to add them to + * @param selectionGroup + * current selection group (or null if none) + */ + public static void addReferenceAnnotations( + Map> annotations, + final AlignmentI alignment, final SequenceGroup selectionGroup) + { + for (SequenceI seq : annotations.keySet()) + { + for (AlignmentAnnotation ann : annotations.get(seq)) + { + AlignmentAnnotation copyAnn = new AlignmentAnnotation(ann); + int startRes = 0; + int endRes = ann.annotations.length; + if (selectionGroup != null) + { + startRes = selectionGroup.getStartRes(); + endRes = selectionGroup.getEndRes(); + } + copyAnn.restrict(startRes, endRes); + + /* + * Add to the sequence (sets copyAnn.datasetSequence), unless the + * original annotation is already on the sequence. + */ + if (!seq.hasAnnotation(ann)) + { + seq.addAlignmentAnnotation(copyAnn); + } + // adjust for gaps + copyAnn.adjustForAlignment(); + // add to the alignment and set visible + alignment.addAnnotation(copyAnn); + copyAnn.visible = true; + } + } + } + + /** + * Set visibility of alignment annotations of specified types (labels), for + * specified sequences. This supports controls like + * "Show all secondary structure", "Hide all Temp factor", etc. + * + * @al the alignment to scan for annotations + * @param types + * the types (labels) of annotations to be updated + * @param forSequences + * if not null, only annotations linked to one of these sequences are + * in scope for update; if null, acts on all sequence annotations + * @param anyType + * if this flag is true, 'types' is ignored (label not checked) + * @param doShow + * if true, set visibility on, else set off + */ + public static void showOrHideSequenceAnnotations(AlignmentI al, + Collection types, List forSequences, + boolean anyType, boolean doShow) + { + for (AlignmentAnnotation aa : al + .getAlignmentAnnotation()) + { + if (anyType || types.contains(aa.label)) + { + if ((aa.sequenceRef != null) + && (forSequences == null || forSequences + .contains(aa.sequenceRef))) + { + aa.visible = doShow; + } + } + } + } }