/* * Jalview - A Sequence Alignment Editor and Viewer ($$Version-Rel$$) * Copyright (C) $$Year-Rel$$ The Jalview Authors * * This file is part of Jalview. * * Jalview is free software: you can redistribute it and/or * modify it under the terms of the GNU General Public License * as published by the Free Software Foundation, either version 3 * of the License, or (at your option) any later version. * * Jalview is distributed in the hope that it will be useful, but * WITHOUT ANY WARRANTY; without even the implied warranty * of MERCHANTABILITY or FITNESS FOR A PARTICULAR * PURPOSE. See the GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with Jalview. If not, see . * The Jalview Authors are detailed in the 'AUTHORS' file. */ package jalview.analysis; import static jalview.io.gff.GffConstants.CLINICAL_SIGNIFICANCE; import jalview.datamodel.AlignedCodon; import jalview.datamodel.AlignedCodonFrame; import jalview.datamodel.AlignedCodonFrame.SequenceToSequenceMapping; import jalview.datamodel.Alignment; import jalview.datamodel.AlignmentAnnotation; import jalview.datamodel.AlignmentI; import jalview.datamodel.DBRefEntry; import jalview.datamodel.GeneLociI; import jalview.datamodel.IncompleteCodonException; import jalview.datamodel.Mapping; import jalview.datamodel.Sequence; import jalview.datamodel.SequenceFeature; import jalview.datamodel.SequenceGroup; import jalview.datamodel.SequenceI; import jalview.datamodel.features.SequenceFeatures; import jalview.io.gff.Gff3Helper; import jalview.io.gff.SequenceOntologyI; import jalview.schemes.ResidueProperties; import jalview.util.Comparison; import jalview.util.DBRefUtils; import jalview.util.IntRangeComparator; import jalview.util.MapList; import jalview.util.MappingUtils; import jalview.util.StringUtils; import java.io.UnsupportedEncodingException; import java.net.URLEncoder; import java.util.ArrayList; import java.util.Arrays; import java.util.Collection; import java.util.Collections; import java.util.HashMap; import java.util.HashSet; import java.util.Iterator; import java.util.LinkedHashMap; import java.util.List; import java.util.Map; import java.util.Map.Entry; import java.util.NoSuchElementException; import java.util.Set; import java.util.SortedMap; import java.util.TreeMap; /** * grab bag of useful alignment manipulation operations Expect these to be * refactored elsewhere at some point. * * @author jimp * */ public class AlignmentUtils { private static final int CODON_LENGTH = 3; private static final String SEQUENCE_VARIANT = "sequence_variant:"; /* * the 'id' attribute is provided for variant features fetched from * Ensembl using its REST service with JSON format */ public static final String VARIANT_ID = "id"; /** * A data model to hold the 'normal' base value at a position, and an optional * sequence variant feature */ static final class DnaVariant { final String base; SequenceFeature variant; DnaVariant(String nuc) { base = nuc; variant = null; } DnaVariant(String nuc, SequenceFeature var) { base = nuc; variant = var; } public String getSource() { return variant == null ? null : variant.getFeatureGroup(); } /** * toString for aid in the debugger only */ @Override public String toString() { return base + ":" + (variant == null ? "" : variant.getDescription()); } } /** * given an existing alignment, create a new alignment including all, or up to * flankSize additional symbols from each sequence's dataset sequence * * @param core * @param flankSize * @return AlignmentI */ public static AlignmentI expandContext(AlignmentI core, int flankSize) { List sq = new ArrayList<>(); int maxoffset = 0; for (SequenceI s : core.getSequences()) { SequenceI newSeq = s.deriveSequence(); final int newSeqStart = newSeq.getStart() - 1; if (newSeqStart > maxoffset && newSeq.getDatasetSequence().getStart() < s.getStart()) { maxoffset = newSeqStart; } sq.add(newSeq); } if (flankSize > -1) { maxoffset = Math.min(maxoffset, flankSize); } /* * now add offset left and right to create an expanded alignment */ for (SequenceI s : sq) { SequenceI ds = s; while (ds.getDatasetSequence() != null) { ds = ds.getDatasetSequence(); } int s_end = s.findPosition(s.getStart() + s.getLength()); // find available flanking residues for sequence int ustream_ds = s.getStart() - ds.getStart(); int dstream_ds = ds.getEnd() - s_end; // build new flanked sequence // compute gap padding to start of flanking sequence int offset = maxoffset - ustream_ds; // padding is gapChar x ( maxoffset - min(ustream_ds, flank) if (flankSize >= 0) { if (flankSize < ustream_ds) { // take up to flankSize residues offset = maxoffset - flankSize; ustream_ds = flankSize; } if (flankSize <= dstream_ds) { dstream_ds = flankSize - 1; } } // TODO use Character.toLowerCase to avoid creating String objects? char[] upstream = new String(ds .getSequence(s.getStart() - 1 - ustream_ds, s.getStart() - 1)) .toLowerCase().toCharArray(); char[] downstream = new String( ds.getSequence(s_end - 1, s_end + dstream_ds)).toLowerCase() .toCharArray(); char[] coreseq = s.getSequence(); char[] nseq = new char[offset + upstream.length + downstream.length + coreseq.length]; char c = core.getGapCharacter(); int p = 0; for (; p < offset; p++) { nseq[p] = c; } System.arraycopy(upstream, 0, nseq, p, upstream.length); System.arraycopy(coreseq, 0, nseq, p + upstream.length, coreseq.length); System.arraycopy(downstream, 0, nseq, p + coreseq.length + upstream.length, downstream.length); s.setSequence(new String(nseq)); s.setStart(s.getStart() - ustream_ds); s.setEnd(s_end + downstream.length); } AlignmentI newAl = new jalview.datamodel.Alignment( sq.toArray(new SequenceI[0])); for (SequenceI s : sq) { if (s.getAnnotation() != null) { for (AlignmentAnnotation aa : s.getAnnotation()) { aa.adjustForAlignment(); // JAL-1712 fix newAl.addAnnotation(aa); } } } newAl.setDataset(core.getDataset()); return newAl; } /** * Returns the index (zero-based position) of a sequence in an alignment, or * -1 if not found. * * @param al * @param seq * @return */ public static int getSequenceIndex(AlignmentI al, SequenceI seq) { int result = -1; int pos = 0; for (SequenceI alSeq : al.getSequences()) { if (alSeq == seq) { result = pos; break; } pos++; } return result; } /** * Returns a map of lists of sequences in the alignment, keyed by sequence * name. For use in mapping between different alignment views of the same * sequences. * * @see jalview.datamodel.AlignmentI#getSequencesByName() */ public static Map> getSequencesByName( AlignmentI al) { Map> theMap = new LinkedHashMap<>(); for (SequenceI seq : al.getSequences()) { String name = seq.getName(); if (name != null) { List seqs = theMap.get(name); if (seqs == null) { seqs = new ArrayList<>(); theMap.put(name, seqs); } seqs.add(seq); } } return theMap; } /** * Build mapping of protein to cDNA alignment. Mappings are made between * 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 boolean mapProteinAlignmentToCdna( final AlignmentI proteinAlignment, final AlignmentI cdnaAlignment) { if (proteinAlignment == null || cdnaAlignment == null) { return false; } Set mappedDna = new HashSet<>(); Set mappedProtein = new HashSet<>(); /* * First pass - map sequences where cross-references exist. This include * 1-to-many mappings to support, for example, variant cDNA. */ 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 mappingExistsOrAdded = false; List thisSeqs = proteinAlignment.getSequences(); for (SequenceI aaSeq : thisSeqs) { boolean proteinMapped = false; AlignedCodonFrame acf = new AlignedCodonFrame(); for (SequenceI cdnaSeq : cdnaAlignment.getSequences()) { /* * 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. */ if (xrefsOnly && !AlignmentUtils.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())) { mappingExistsOrAdded = true; } else { MapList map = mapCdnaToProtein(aaSeq, cdnaSeq); if (map != null) { acf.addMap(cdnaSeq, aaSeq, map); mappingExistsOrAdded = true; proteinMapped = true; mappedDna.add(cdnaSeq); mappedProtein.add(aaSeq); } } } if (proteinMapped) { proteinAlignment.addCodonFrame(acf); } } return mappingExistsOrAdded; } /** * Answers true if the mappings include one between the given (dataset) * sequences. */ protected static boolean mappingExists(List mappings, SequenceI aaSeq, SequenceI cdnaSeq) { if (mappings != null) { for (AlignedCodonFrame acf : mappings) { if (cdnaSeq == acf.getDnaForAaSeq(aaSeq)) { return true; } } } return false; } /** * Builds a mapping (if possible) of a cDNA to a protein sequence. *
    *
  • first checks if the cdna translates exactly to the protein * sequence
  • *
  • else checks for translation after removing a STOP codon
  • *
  • else checks for translation after removing a START codon
  • *
  • if that fails, inspect CDS features on the cDNA sequence
  • *
* Returns null if no mapping is determined. * * @param proteinSeq * the aligned protein sequence * @param cdnaSeq * the aligned cdna sequence * @return */ public static MapList mapCdnaToProtein(SequenceI proteinSeq, SequenceI cdnaSeq) { /* * 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(); char[] aaSeqChars = proteinDataset != null ? proteinDataset.getSequence() : proteinSeq.getSequence(); final SequenceI cdnaDataset = cdnaSeq.getDatasetSequence(); char[] cdnaSeqChars = cdnaDataset != null ? cdnaDataset.getSequence() : cdnaSeq.getSequence(); if (aaSeqChars == null || cdnaSeqChars == null) { return null; } /* * cdnaStart/End, proteinStartEnd are base 1 (for dataset sequence mapping) */ final int mappedLength = CODON_LENGTH * aaSeqChars.length; int cdnaLength = cdnaSeqChars.length; int cdnaStart = cdnaSeq.getStart(); int cdnaEnd = cdnaSeq.getEnd(); final int proteinStart = proteinSeq.getStart(); final int proteinEnd = proteinSeq.getEnd(); /* * If lengths don't match, try ignoring stop codon (if present) */ if (cdnaLength != mappedLength && cdnaLength > 2) { String lastCodon = String.valueOf(cdnaSeqChars, cdnaLength - CODON_LENGTH, CODON_LENGTH).toUpperCase(); for (String stop : ResidueProperties.STOP_CODONS) { if (lastCodon.equals(stop)) { cdnaEnd -= CODON_LENGTH; cdnaLength -= CODON_LENGTH; break; } } } /* * If lengths still don't match, try ignoring start codon. */ int startOffset = 0; if (cdnaLength != mappedLength && cdnaLength > 2 && String.valueOf(cdnaSeqChars, 0, CODON_LENGTH).toUpperCase() .equals(ResidueProperties.START)) { startOffset += CODON_LENGTH; cdnaStart += CODON_LENGTH; cdnaLength -= CODON_LENGTH; } if (translatesAs(cdnaSeqChars, startOffset, aaSeqChars)) { /* * protein is translation of dna (+/- start/stop codons) */ MapList map = new MapList(new int[] { cdnaStart, cdnaEnd }, new int[] { proteinStart, proteinEnd }, CODON_LENGTH, 1); return map; } /* * translation failed - try mapping CDS annotated regions of dna */ return mapCdsToProtein(cdnaSeq, proteinSeq); } /** * 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) { if (cdnaSeqChars == null || aaSeqChars == null) { return false; } int aaPos = 0; int dnaPos = cdnaStart; for (; dnaPos < cdnaSeqChars.length - 2 && aaPos < aaSeqChars.length; dnaPos += CODON_LENGTH, aaPos++) { String codon = String.valueOf(cdnaSeqChars, dnaPos, CODON_LENGTH); final String translated = ResidueProperties.codonTranslate(codon); /* * allow * in protein to match untranslatable in dna */ final char aaRes = aaSeqChars[aaPos]; if ((translated == null || ResidueProperties.STOP.equals(translated)) && aaRes == '*') { continue; } if (translated == null || !(aaRes == translated.charAt(0))) { // debug // System.out.println(("Mismatch at " + i + "/" + aaResidue + ": " // + codon + "(" + translated + ") != " + aaRes)); return false; } } /* * check we matched all of the protein sequence */ if (aaPos != aaSeqChars.length) { return false; } /* * check we matched all of the dna except * for optional trailing STOP codon */ if (dnaPos == cdnaSeqChars.length) { return true; } if (dnaPos == cdnaSeqChars.length - CODON_LENGTH) { String codon = String.valueOf(cdnaSeqChars, dnaPos, CODON_LENGTH); if (ResidueProperties.STOP .equals(ResidueProperties.codonTranslate(codon))) { return true; } } return false; } /** * Align sequence 'seq' to match the alignment of a mapped sequence. Note this * currently assumes that we are aligning cDNA to match protein. * * @param seq * the sequence to be realigned * @param al * the alignment whose sequence alignment is to be 'copied' * @param gap * character string represent a gap in the realigned sequence * @param preserveUnmappedGaps * @param preserveMappedGaps * @return true if the sequence was realigned, false if it could not be */ public static boolean alignSequenceAs(SequenceI seq, AlignmentI al, String gap, boolean preserveMappedGaps, boolean preserveUnmappedGaps) { /* * Get any mappings from the source alignment to the target (dataset) * sequence. */ // 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 || mappings.isEmpty()) { return false; } /* * Locate the aligned source sequence whose dataset sequence is mapped. We * just take the first match here (as we can't align like more than one * sequence). */ SequenceI alignFrom = null; AlignedCodonFrame mapping = null; for (AlignedCodonFrame mp : mappings) { alignFrom = mp.findAlignedSequence(seq, al); if (alignFrom != null) { mapping = mp; break; } } if (alignFrom == null) { return false; } alignSequenceAs(seq, alignFrom, mapping, gap, al.getGapCharacter(), preserveMappedGaps, preserveUnmappedGaps); return true; } /** * Align sequence 'alignTo' the same way as 'alignFrom', using the mapping to * match residues and codons. Flags control whether existing gaps in unmapped * (intron) and mapped (exon) regions are preserved or not. Gaps between * intron and exon are only retained if both flags are set. * * @param alignTo * @param alignFrom * @param mapping * @param myGap * @param sourceGap * @param preserveUnmappedGaps * @param preserveMappedGaps */ public static void alignSequenceAs(SequenceI alignTo, SequenceI alignFrom, AlignedCodonFrame mapping, String myGap, char sourceGap, boolean preserveMappedGaps, boolean preserveUnmappedGaps) { // TODO generalise to work for Protein-Protein, dna-dna, dna-protein // aligned and dataset sequence positions, all base zero int thisSeqPos = 0; int sourceDsPos = 0; int basesWritten = 0; char myGapChar = myGap.charAt(0); int ratio = myGap.length(); int fromOffset = alignFrom.getStart() - 1; int toOffset = alignTo.getStart() - 1; int sourceGapMappedLength = 0; boolean inExon = false; final int toLength = alignTo.getLength(); final int fromLength = alignFrom.getLength(); StringBuilder thisAligned = new StringBuilder(2 * toLength); /* * Traverse the 'model' aligned sequence */ for (int i = 0; i < fromLength; i++) { char sourceChar = alignFrom.getCharAt(i); if (sourceChar == sourceGap) { sourceGapMappedLength += ratio; continue; } /* * Found a non-gap character. Locate its mapped region if any. */ sourceDsPos++; // Note mapping positions are base 1, our sequence positions base 0 int[] mappedPos = mapping.getMappedRegion(alignTo, alignFrom, sourceDsPos + fromOffset); if (mappedPos == null) { /* * unmapped position; treat like a gap */ sourceGapMappedLength += ratio; // System.err.println("Can't align: no codon mapping to residue " // + sourceDsPos + "(" + sourceChar + ")"); // return; continue; } int mappedCodonStart = mappedPos[0]; // position (1...) of codon start int mappedCodonEnd = mappedPos[mappedPos.length - 1]; // codon end pos StringBuilder trailingCopiedGap = new StringBuilder(); /* * Copy dna sequence up to and including this codon. Optionally, include * gaps before the codon starts (in introns) and/or after the codon starts * (in exons). * * Note this only works for 'linear' splicing, not reverse or interleaved. * But then 'align dna as protein' doesn't make much sense otherwise. */ int intronLength = 0; while (basesWritten + toOffset < mappedCodonEnd && thisSeqPos < toLength) { final char c = alignTo.getCharAt(thisSeqPos++); if (c != myGapChar) { basesWritten++; int sourcePosition = basesWritten + toOffset; if (sourcePosition < mappedCodonStart) { /* * Found an unmapped (intron) base. First add in any preceding gaps * (if wanted). */ if (preserveUnmappedGaps && trailingCopiedGap.length() > 0) { thisAligned.append(trailingCopiedGap.toString()); intronLength += trailingCopiedGap.length(); trailingCopiedGap = new StringBuilder(); } intronLength++; inExon = false; } else { final boolean startOfCodon = sourcePosition == mappedCodonStart; int gapsToAdd = calculateGapsToInsert(preserveMappedGaps, preserveUnmappedGaps, sourceGapMappedLength, inExon, trailingCopiedGap.length(), intronLength, startOfCodon); for (int k = 0; k < gapsToAdd; k++) { thisAligned.append(myGapChar); } sourceGapMappedLength = 0; inExon = true; } thisAligned.append(c); trailingCopiedGap = new StringBuilder(); } else { if (inExon && preserveMappedGaps) { trailingCopiedGap.append(myGapChar); } else if (!inExon && preserveUnmappedGaps) { trailingCopiedGap.append(myGapChar); } } } } /* * At end of model aligned sequence. Copy any remaining target sequence, optionally * including (intron) gaps. */ while (thisSeqPos < toLength) { final char c = alignTo.getCharAt(thisSeqPos++); if (c != myGapChar || preserveUnmappedGaps) { thisAligned.append(c); } sourceGapMappedLength--; } /* * finally add gaps to pad for any trailing source gaps or * unmapped characters */ if (preserveUnmappedGaps) { while (sourceGapMappedLength > 0) { thisAligned.append(myGapChar); sourceGapMappedLength--; } } /* * All done aligning, set the aligned sequence. */ alignTo.setSequence(new String(thisAligned)); } /** * Helper method to work out how many gaps to insert when realigning. * * @param preserveMappedGaps * @param preserveUnmappedGaps * @param sourceGapMappedLength * @param inExon * @param trailingCopiedGap * @param intronLength * @param startOfCodon * @return */ protected static int calculateGapsToInsert(boolean preserveMappedGaps, boolean preserveUnmappedGaps, int sourceGapMappedLength, boolean inExon, int trailingGapLength, int intronLength, final boolean startOfCodon) { int gapsToAdd = 0; if (startOfCodon) { /* * Reached start of codon. Ignore trailing gaps in intron unless we are * preserving gaps in both exon and intron. Ignore them anyway if the * protein alignment introduces a gap at least as large as the intronic * region. */ if (inExon && !preserveMappedGaps) { trailingGapLength = 0; } if (!inExon && !(preserveMappedGaps && preserveUnmappedGaps)) { trailingGapLength = 0; } if (inExon) { gapsToAdd = Math.max(sourceGapMappedLength, trailingGapLength); } else { if (intronLength + trailingGapLength <= sourceGapMappedLength) { gapsToAdd = sourceGapMappedLength - intronLength; } else { gapsToAdd = Math.min( intronLength + trailingGapLength - sourceGapMappedLength, trailingGapLength); } } } else { /* * second or third base of codon; check for any gaps in dna */ if (!preserveMappedGaps) { trailingGapLength = 0; } gapsToAdd = Math.max(sourceGapMappedLength, trailingGapLength); } return gapsToAdd; } /** * Realigns the given protein to match the alignment of the dna, using codon * mappings to translate aligned codon positions to protein residues. * * @param protein * the alignment whose sequences are realigned by this method * @param dna * the dna alignment whose alignment we are 'copying' * @return the number of sequences that were realigned */ public static int alignProteinAsDna(AlignmentI protein, AlignmentI dna) { if (protein.isNucleotide() || !dna.isNucleotide()) { System.err.println("Wrong alignment type in alignProteinAsDna"); return 0; } List unmappedProtein = new ArrayList<>(); Map> alignedCodons = buildCodonColumnsMap( protein, dna, unmappedProtein); return alignProteinAs(protein, alignedCodons, unmappedProtein); } /** * Realigns the given dna to match the alignment of the protein, using codon * mappings to translate aligned peptide positions to codons. * * Always produces a padded CDS alignment. * * @param dna * the alignment whose sequences are realigned by this method * @param protein * the protein alignment whose alignment we are 'copying' * @return the number of sequences that were realigned */ public static int alignCdsAsProtein(AlignmentI dna, AlignmentI protein) { if (protein.isNucleotide() || !dna.isNucleotide()) { System.err.println("Wrong alignment type in alignProteinAsDna"); return 0; } // todo: implement this List mappings = protein.getCodonFrames(); int alignedCount = 0; int width = 0; // alignment width for padding CDS for (SequenceI dnaSeq : dna.getSequences()) { if (alignCdsSequenceAsProtein(dnaSeq, protein, mappings, dna.getGapCharacter())) { alignedCount++; } width = Math.max(dnaSeq.getLength(), width); } int oldwidth; int diff; for (SequenceI dnaSeq : dna.getSequences()) { oldwidth = dnaSeq.getLength(); diff = width - oldwidth; if (diff > 0) { dnaSeq.insertCharAt(oldwidth, diff, dna.getGapCharacter()); } } return alignedCount; } /** * Helper method to align (if possible) the dna sequence to match the * alignment of a mapped protein sequence. This is currently limited to * handling coding sequence only. * * @param cdsSeq * @param protein * @param mappings * @param gapChar * @return */ static boolean alignCdsSequenceAsProtein(SequenceI cdsSeq, AlignmentI protein, List mappings, char gapChar) { SequenceI cdsDss = cdsSeq.getDatasetSequence(); if (cdsDss == null) { System.err .println("alignCdsSequenceAsProtein needs aligned sequence!"); return false; } List dnaMappings = MappingUtils .findMappingsForSequence(cdsSeq, mappings); for (AlignedCodonFrame mapping : dnaMappings) { SequenceI peptide = mapping.findAlignedSequence(cdsSeq, protein); if (peptide != null) { final int peptideLength = peptide.getLength(); Mapping map = mapping.getMappingBetween(cdsSeq, peptide); if (map != null) { MapList mapList = map.getMap(); if (map.getTo() == peptide.getDatasetSequence()) { mapList = mapList.getInverse(); } final int cdsLength = cdsDss.getLength(); int mappedFromLength = MappingUtils.getLength(mapList .getFromRanges()); int mappedToLength = MappingUtils .getLength(mapList.getToRanges()); boolean addStopCodon = (cdsLength == mappedFromLength * CODON_LENGTH + CODON_LENGTH) || (peptide.getDatasetSequence() .getLength() == mappedFromLength - 1); if (cdsLength != mappedToLength && !addStopCodon) { System.err.println(String.format( "Can't align cds as protein (length mismatch %d/%d): %s", cdsLength, mappedToLength, cdsSeq.getName())); } /* * pre-fill the aligned cds sequence with gaps */ char[] alignedCds = new char[peptideLength * CODON_LENGTH + (addStopCodon ? CODON_LENGTH : 0)]; Arrays.fill(alignedCds, gapChar); /* * walk over the aligned peptide sequence and insert mapped * codons for residues in the aligned cds sequence */ int copiedBases = 0; int cdsStart = cdsDss.getStart(); int proteinPos = peptide.getStart() - 1; int cdsCol = 0; for (int col = 0; col < peptideLength; col++) { char residue = peptide.getCharAt(col); if (Comparison.isGap(residue)) { cdsCol += CODON_LENGTH; } else { proteinPos++; int[] codon = mapList.locateInTo(proteinPos, proteinPos); if (codon == null) { // e.g. incomplete start codon, X in peptide cdsCol += CODON_LENGTH; } else { for (int j = codon[0]; j <= codon[1]; j++) { char mappedBase = cdsDss.getCharAt(j - cdsStart); alignedCds[cdsCol++] = mappedBase; copiedBases++; } } } } /* * append stop codon if not mapped from protein, * closing it up to the end of the mapped sequence */ if (copiedBases == cdsLength - CODON_LENGTH) { for (int i = alignedCds.length - 1; i >= 0; i--) { if (!Comparison.isGap(alignedCds[i])) { cdsCol = i + 1; // gap just after end of sequence break; } } for (int i = cdsLength - CODON_LENGTH; i < cdsLength; i++) { alignedCds[cdsCol++] = cdsDss.getCharAt(i); } } cdsSeq.setSequence(new String(alignedCds)); return true; } } } return false; } /** * Builds a map whose key is an aligned codon position (3 alignment column * numbers base 0), and whose value is a map from protein sequence to each * protein's peptide residue for that codon. The map generates an ordering of * the codons, and allows us to read off the peptides at each position in * order to assemble 'aligned' protein sequences. * * @param protein * the protein alignment * @param dna * the coding dna alignment * @param unmappedProtein * any unmapped proteins are added to this list * @return */ protected static Map> buildCodonColumnsMap( AlignmentI protein, AlignmentI dna, List unmappedProtein) { /* * maintain a list of any proteins with no mappings - these will be * rendered 'as is' in the protein alignment as we can't align them */ unmappedProtein.addAll(protein.getSequences()); List mappings = protein.getCodonFrames(); /* * Map will hold, for each aligned codon position e.g. [3, 5, 6], a map of * {dnaSequence, {proteinSequence, codonProduct}} at that position. The * comparator keeps the codon positions ordered. */ Map> alignedCodons = new TreeMap<>( new CodonComparator()); for (SequenceI dnaSeq : dna.getSequences()) { for (AlignedCodonFrame mapping : mappings) { SequenceI prot = mapping.findAlignedSequence(dnaSeq, protein); if (prot != null) { Mapping seqMap = mapping.getMappingForSequence(dnaSeq); addCodonPositions(dnaSeq, prot, protein.getGapCharacter(), seqMap, alignedCodons); unmappedProtein.remove(prot); } } } /* * Finally add any unmapped peptide start residues (e.g. for incomplete * codons) as if at the codon position before the second residue */ // TODO resolve JAL-2022 so this fudge can be removed int mappedSequenceCount = protein.getHeight() - unmappedProtein.size(); addUnmappedPeptideStarts(alignedCodons, mappedSequenceCount); return alignedCodons; } /** * Scans for any protein mapped from position 2 (meaning unmapped start * position e.g. an incomplete codon), and synthesizes a 'codon' for it at the * preceding position in the alignment * * @param alignedCodons * the codon-to-peptide map * @param mappedSequenceCount * the number of distinct sequences in the map */ protected static void addUnmappedPeptideStarts( Map> alignedCodons, int mappedSequenceCount) { // TODO delete this ugly hack once JAL-2022 is resolved // i.e. we can model startPhase > 0 (incomplete start codon) List sequencesChecked = new ArrayList<>(); AlignedCodon lastCodon = null; Map toAdd = new HashMap<>(); for (Entry> entry : alignedCodons .entrySet()) { for (Entry sequenceCodon : entry.getValue() .entrySet()) { SequenceI seq = sequenceCodon.getKey(); if (sequencesChecked.contains(seq)) { continue; } sequencesChecked.add(seq); AlignedCodon codon = sequenceCodon.getValue(); if (codon.peptideCol > 1) { System.err.println( "Problem mapping protein with >1 unmapped start positions: " + seq.getName()); } else if (codon.peptideCol == 1) { /* * first position (peptideCol == 0) was unmapped - add it */ if (lastCodon != null) { AlignedCodon firstPeptide = new AlignedCodon(lastCodon.pos1, lastCodon.pos2, lastCodon.pos3, String.valueOf(seq.getCharAt(0)), 0); toAdd.put(seq, firstPeptide); } else { /* * unmapped residue at start of alignment (no prior column) - * 'insert' at nominal codon [0, 0, 0] */ AlignedCodon firstPeptide = new AlignedCodon(0, 0, 0, String.valueOf(seq.getCharAt(0)), 0); toAdd.put(seq, firstPeptide); } } if (sequencesChecked.size() == mappedSequenceCount) { // no need to check past first mapped position in all sequences break; } } lastCodon = entry.getKey(); } /* * add any new codons safely after iterating over the map */ for (Entry startCodon : toAdd.entrySet()) { addCodonToMap(alignedCodons, startCodon.getValue(), startCodon.getKey()); } } /** * Update the aligned protein sequences to match the codon alignments given in * the map. * * @param protein * @param alignedCodons * an ordered map of codon positions (columns), with sequence/peptide * values present in each column * @param unmappedProtein * @return */ protected static int alignProteinAs(AlignmentI protein, Map> alignedCodons, List unmappedProtein) { /* * prefill peptide sequences with gaps */ int alignedWidth = alignedCodons.size(); char[] gaps = new char[alignedWidth]; Arrays.fill(gaps, protein.getGapCharacter()); Map peptides = new HashMap<>(); for (SequenceI seq : protein.getSequences()) { if (!unmappedProtein.contains(seq)) { peptides.put(seq, Arrays.copyOf(gaps, gaps.length)); } } /* * Traverse the codons left to right (as defined by CodonComparator) * and insert peptides in each column where the sequence is mapped. * This gives a peptide 'alignment' where residues are aligned if their * corresponding codons occupy the same columns in the cdna alignment. */ int column = 0; for (AlignedCodon codon : alignedCodons.keySet()) { final Map columnResidues = alignedCodons .get(codon); for (Entry entry : columnResidues.entrySet()) { char residue = entry.getValue().product.charAt(0); peptides.get(entry.getKey())[column] = residue; } column++; } /* * and finally set the constructed sequences */ for (Entry entry : peptides.entrySet()) { entry.getKey().setSequence(new String(entry.getValue())); } return 0; } /** * Populate the map of aligned codons by traversing the given sequence * mapping, locating the aligned positions of mapped codons, and adding those * positions and their translation products to the map. * * @param dna * the aligned sequence we are mapping from * @param protein * the sequence to be aligned to the codons * @param gapChar * the gap character in the dna sequence * @param seqMap * a mapping to a sequence translation * @param alignedCodons * the map we are building up */ static void addCodonPositions(SequenceI dna, SequenceI protein, char gapChar, Mapping seqMap, Map> alignedCodons) { Iterator codons = seqMap.getCodonIterator(dna, gapChar); /* * add codon positions, and their peptide translations, to the alignment * map, while remembering the first codon mapped */ while (codons.hasNext()) { try { AlignedCodon codon = codons.next(); addCodonToMap(alignedCodons, codon, protein); } catch (IncompleteCodonException e) { // possible incomplete trailing codon - ignore } catch (NoSuchElementException e) { // possibly peptide lacking STOP } } } /** * Helper method to add a codon-to-peptide entry to the aligned codons map * * @param alignedCodons * @param codon * @param protein */ protected static void addCodonToMap( Map> alignedCodons, AlignedCodon codon, SequenceI protein) { Map seqProduct = alignedCodons.get(codon); if (seqProduct == null) { seqProduct = new HashMap<>(); alignedCodons.put(codon, seqProduct); } seqProduct.put(protein, codon); } /** * 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) { if (al1 == null || al2 == null) { return false; } /* * 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; List 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 */ protected static boolean isMappable(SequenceI dnaSeq, SequenceI proteinSeq, List mappings) { if (dnaSeq == null || proteinSeq == null) { return false; } SequenceI dnaDs = dnaSeq.getDatasetSequence() == null ? dnaSeq : dnaSeq.getDatasetSequence(); SequenceI proteinDs = proteinSeq.getDatasetSequence() == null ? proteinSeq : proteinSeq.getDatasetSequence(); for (AlignedCodonFrame mapping : mappings) { if (proteinDs == mapping.getAaForDnaSeq(dnaDs)) { /* * already mapped */ return true; } } /* * Just try to make a mapping (it is not yet stored), test whether * successful. */ return mapCdnaToProtein(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) { AlignmentAnnotation[] anns = al.getAlignmentAnnotation(); if (anns != null) { for (AlignmentAnnotation aa : anns) { if (anyType || types.contains(aa.label)) { if ((aa.sequenceRef != null) && (forSequences == null || forSequences.contains(aa.sequenceRef))) { aa.visible = doShow; } } } } } /** * Returns true if either sequence has a cross-reference to the other * * @param seq1 * @param seq2 * @return */ public static boolean haveCrossRef(SequenceI seq1, SequenceI seq2) { // Note: moved here from class CrossRef as the latter class has dependencies // not availability to the applet's classpath return hasCrossRef(seq1, seq2) || hasCrossRef(seq2, seq1); } /** * Returns true if seq1 has a cross-reference to seq2. Currently this assumes * that sequence name is structured as Source|AccessionId. * * @param seq1 * @param seq2 * @return */ public static boolean hasCrossRef(SequenceI seq1, SequenceI seq2) { if (seq1 == null || seq2 == null) { return false; } String name = seq2.getName(); final DBRefEntry[] xrefs = seq1.getDBRefs(); if (xrefs != null) { for (DBRefEntry xref : xrefs) { String xrefName = xref.getSource() + "|" + xref.getAccessionId(); // case-insensitive test, consistent with DBRefEntry.equalRef() if (xrefName.equalsIgnoreCase(name)) { return true; } } } return false; } /** * Constructs an alignment consisting of the mapped (CDS) regions in the given * nucleotide sequences, and updates mappings to match. The CDS sequences are * added to the original alignment's dataset, which is shared by the new * alignment. Mappings from nucleotide to CDS, and from CDS to protein, are * added to the alignment dataset. * * @param dna * aligned nucleotide (dna or cds) sequences * @param dataset * the alignment dataset the sequences belong to * @param products * (optional) to restrict results to CDS that map to specified * protein products * @return an alignment whose sequences are the cds-only parts of the dna * sequences (or null if no mappings are found) */ public static AlignmentI makeCdsAlignment(SequenceI[] dna, AlignmentI dataset, SequenceI[] products) { if (dataset == null || dataset.getDataset() != null) { throw new IllegalArgumentException( "IMPLEMENTATION ERROR: dataset.getDataset() must be null!"); } List foundSeqs = new ArrayList<>(); List cdsSeqs = new ArrayList<>(); List mappings = dataset.getCodonFrames(); HashSet productSeqs = null; if (products != null) { productSeqs = new HashSet<>(); for (SequenceI seq : products) { productSeqs.add(seq.getDatasetSequence() == null ? seq : seq .getDatasetSequence()); } } /* * Construct CDS sequences from mappings on the alignment dataset. * The logic is: * - find the protein product(s) mapped to from each dna sequence * - if the mapping covers the whole dna sequence (give or take start/stop * codon), take the dna as the CDS sequence * - else search dataset mappings for a suitable dna sequence, i.e. one * whose whole sequence is mapped to the protein * - if no sequence found, construct one from the dna sequence and mapping * (and add it to dataset so it is found if this is repeated) */ for (SequenceI dnaSeq : dna) { SequenceI dnaDss = dnaSeq.getDatasetSequence() == null ? dnaSeq : dnaSeq.getDatasetSequence(); List seqMappings = MappingUtils .findMappingsForSequence(dnaSeq, mappings); for (AlignedCodonFrame mapping : seqMappings) { List mappingsFromSequence = mapping .getMappingsFromSequence(dnaSeq); for (Mapping aMapping : mappingsFromSequence) { MapList mapList = aMapping.getMap(); if (mapList.getFromRatio() == 1) { /* * not a dna-to-protein mapping (likely dna-to-cds) */ continue; } /* * skip if mapping is not to one of the target set of proteins */ SequenceI proteinProduct = aMapping.getTo(); if (productSeqs != null && !productSeqs.contains(proteinProduct)) { continue; } /* * try to locate the CDS from the dataset mappings; * guard against duplicate results (for the case that protein has * dbrefs to both dna and cds sequences) */ SequenceI cdsSeq = findCdsForProtein(mappings, dnaSeq, seqMappings, aMapping); if (cdsSeq != null) { if (!foundSeqs.contains(cdsSeq)) { foundSeqs.add(cdsSeq); SequenceI derivedSequence = cdsSeq.deriveSequence(); cdsSeqs.add(derivedSequence); if (!dataset.getSequences().contains(cdsSeq)) { dataset.addSequence(cdsSeq); } } continue; } /* * didn't find mapped CDS sequence - construct it and add * its dataset sequence to the dataset */ cdsSeq = makeCdsSequence(dnaSeq.getDatasetSequence(), aMapping, dataset).deriveSequence(); // cdsSeq has a name constructed as CDS| // will be either the accession for the coding sequence, // marked in the /via/ dbref to the protein product accession // or it will be the original nucleotide accession. SequenceI cdsSeqDss = cdsSeq.getDatasetSequence(); cdsSeqs.add(cdsSeq); if (!dataset.getSequences().contains(cdsSeqDss)) { // check if this sequence is a newly created one // so needs adding to the dataset dataset.addSequence(cdsSeqDss); } /* * add a mapping from CDS to the (unchanged) mapped to range */ List cdsRange = Collections.singletonList(new int[] { 1, cdsSeq.getLength() }); MapList cdsToProteinMap = new MapList(cdsRange, mapList.getToRanges(), mapList.getFromRatio(), mapList.getToRatio()); AlignedCodonFrame cdsToProteinMapping = new AlignedCodonFrame(); cdsToProteinMapping.addMap(cdsSeqDss, proteinProduct, cdsToProteinMap); /* * guard against duplicating the mapping if repeating this action */ if (!mappings.contains(cdsToProteinMapping)) { mappings.add(cdsToProteinMapping); } propagateDBRefsToCDS(cdsSeqDss, dnaSeq.getDatasetSequence(), proteinProduct, aMapping); /* * add another mapping from original 'from' range to CDS */ AlignedCodonFrame dnaToCdsMapping = new AlignedCodonFrame(); final MapList dnaToCdsMap = new MapList(mapList.getFromRanges(), cdsRange, 1, 1); dnaToCdsMapping.addMap(dnaSeq.getDatasetSequence(), cdsSeqDss, dnaToCdsMap); if (!mappings.contains(dnaToCdsMapping)) { mappings.add(dnaToCdsMapping); } /* * transfer dna chromosomal loci (if known) to the CDS * sequence (via the mapping) */ final MapList cdsToDnaMap = dnaToCdsMap.getInverse(); transferGeneLoci(dnaSeq, cdsToDnaMap, cdsSeq); /* * add DBRef with mapping from protein to CDS * (this enables Get Cross-References from protein alignment) * This is tricky because we can't have two DBRefs with the * same source and accession, so need a different accession for * the CDS from the dna sequence */ // specific use case: // Genomic contig ENSCHR:1, contains coding regions for ENSG01, // ENSG02, ENSG03, with transcripts and products similarly named. // cannot add distinct dbrefs mapping location on ENSCHR:1 to ENSG01 // JBPNote: ?? can't actually create an example that demonstrates we // need to // synthesize an xref. for (DBRefEntry primRef : dnaDss.getPrimaryDBRefs()) { /* * create a cross-reference from CDS to the source sequence's * primary reference and vice versa */ String source = primRef.getSource(); String version = primRef.getVersion(); DBRefEntry cdsCrossRef = new DBRefEntry(source, source + ":" + version, primRef.getAccessionId()); cdsCrossRef.setMap(new Mapping(dnaDss, new MapList(cdsToDnaMap))); cdsSeqDss.addDBRef(cdsCrossRef); dnaSeq.addDBRef(new DBRefEntry(source, version, cdsSeq .getName(), new Mapping(cdsSeqDss, dnaToCdsMap))); // problem here is that the cross-reference is synthesized - // cdsSeq.getName() may be like 'CDS|dnaaccession' or // 'CDS|emblcdsacc' // assuming cds version same as dna ?!? DBRefEntry proteinToCdsRef = new DBRefEntry(source, version, cdsSeq.getName()); // proteinToCdsRef.setMap(new Mapping(cdsSeqDss, cdsToProteinMap .getInverse())); proteinProduct.addDBRef(proteinToCdsRef); } /* * transfer any features on dna that overlap the CDS */ transferFeatures(dnaSeq, cdsSeq, dnaToCdsMap, null, SequenceOntologyI.CDS); } } } AlignmentI cds = new Alignment(cdsSeqs.toArray(new SequenceI[cdsSeqs .size()])); cds.setDataset(dataset); return cds; } /** * Tries to transfer gene loci (dbref to chromosome positions) from fromSeq to * toSeq, mediated by the given mapping between the sequences * * @param fromSeq * @param targetToFrom * Map * @param targetSeq */ protected static void transferGeneLoci(SequenceI fromSeq, MapList targetToFrom, SequenceI targetSeq) { if (targetSeq.getGeneLoci() != null) { // already have - don't override return; } GeneLociI fromLoci = fromSeq.getGeneLoci(); if (fromLoci == null) { return; } MapList newMap = targetToFrom.traverse(fromLoci.getMap()); if (newMap != null) { targetSeq.setGeneLoci(fromLoci.getSpeciesId(), fromLoci.getAssemblyId(), fromLoci.getChromosomeId(), newMap); } } /** * A helper method that finds a CDS sequence in the alignment dataset that is * mapped to the given protein sequence, and either is, or has a mapping from, * the given dna sequence. * * @param mappings * set of all mappings on the dataset * @param dnaSeq * a dna (or cds) sequence we are searching from * @param seqMappings * the set of mappings involving dnaSeq * @param aMapping * a transcript-to-peptide mapping * @return */ static SequenceI findCdsForProtein(List mappings, SequenceI dnaSeq, List seqMappings, Mapping aMapping) { /* * TODO a better dna-cds-protein mapping data representation to allow easy * navigation; until then this clunky looping around lists of mappings */ SequenceI seqDss = dnaSeq.getDatasetSequence() == null ? dnaSeq : dnaSeq.getDatasetSequence(); SequenceI proteinProduct = aMapping.getTo(); /* * is this mapping from the whole dna sequence (i.e. CDS)? * allowing for possible stop codon on dna but not peptide */ int mappedFromLength = MappingUtils .getLength(aMapping.getMap().getFromRanges()); int dnaLength = seqDss.getLength(); if (mappedFromLength == dnaLength || mappedFromLength == dnaLength - CODON_LENGTH) { /* * if sequence has CDS features, this is a transcript with no UTR * - do not take this as the CDS sequence! (JAL-2789) */ if (seqDss.getFeatures().getFeaturesByOntology(SequenceOntologyI.CDS) .isEmpty()) { return seqDss; } } /* * looks like we found the dna-to-protein mapping; search for the * corresponding cds-to-protein mapping */ List mappingsToPeptide = MappingUtils .findMappingsForSequence(proteinProduct, mappings); for (AlignedCodonFrame acf : mappingsToPeptide) { for (SequenceToSequenceMapping map : acf.getMappings()) { Mapping mapping = map.getMapping(); if (mapping != aMapping && mapping.getMap().getFromRatio() == CODON_LENGTH && proteinProduct == mapping.getTo() && seqDss != map.getFromSeq()) { mappedFromLength = MappingUtils .getLength(mapping.getMap().getFromRanges()); if (mappedFromLength == map.getFromSeq().getLength()) { /* * found a 3:1 mapping to the protein product which covers * the whole dna sequence i.e. is from CDS; finally check the CDS * is mapped from the given dna start sequence */ SequenceI cdsSeq = map.getFromSeq(); // todo this test is weak if seqMappings contains multiple mappings; // we get away with it if transcript:cds relationship is 1:1 List dnaToCdsMaps = MappingUtils .findMappingsForSequence(cdsSeq, seqMappings); if (!dnaToCdsMaps.isEmpty()) { return cdsSeq; } } } } } return null; } /** * Helper method that makes a CDS sequence as defined by the mappings from the * given sequence i.e. extracts the 'mapped from' ranges (which may be on * forward or reverse strand). * * @param seq * @param mapping * @param dataset * - existing dataset. We check for sequences that look like the CDS * we are about to construct, if one exists already, then we will * just return that one. * @return CDS sequence (as a dataset sequence) */ static SequenceI makeCdsSequence(SequenceI seq, Mapping mapping, AlignmentI dataset) { char[] seqChars = seq.getSequence(); List fromRanges = mapping.getMap().getFromRanges(); int cdsWidth = MappingUtils.getLength(fromRanges); char[] newSeqChars = new char[cdsWidth]; int newPos = 0; for (int[] range : fromRanges) { if (range[0] <= range[1]) { // forward strand mapping - just copy the range int length = range[1] - range[0] + 1; System.arraycopy(seqChars, range[0] - 1, newSeqChars, newPos, length); newPos += length; } else { // reverse strand mapping - copy and complement one by one for (int i = range[0]; i >= range[1]; i--) { newSeqChars[newPos++] = Dna.getComplement(seqChars[i - 1]); } } } /* * assign 'from id' held in the mapping if set (e.g. EMBL protein_id), * else generate a sequence name */ String mapFromId = mapping.getMappedFromId(); String seqId = "CDS|" + (mapFromId != null ? mapFromId : seq.getName()); SequenceI newSeq = new Sequence(seqId, newSeqChars, 1, newPos); if (dataset != null) { SequenceI[] matches = dataset.findSequenceMatch(newSeq.getName()); if (matches != null) { boolean matched = false; for (SequenceI mtch : matches) { if (mtch.getStart() != newSeq.getStart()) { continue; } if (mtch.getEnd() != newSeq.getEnd()) { continue; } if (!Arrays.equals(mtch.getSequence(), newSeq.getSequence())) { continue; } if (!matched) { matched = true; newSeq = mtch; } else { System.err.println( "JAL-2154 regression: warning - found (and ignnored a duplicate CDS sequence):" + mtch.toString()); } } } } // newSeq.setDescription(mapFromId); return newSeq; } /** * Adds any DBRefEntrys to cdsSeq from contig that have a Mapping congruent to * the given mapping. * * @param cdsSeq * @param contig * @param proteinProduct * @param mapping * @return list of DBRefEntrys added */ protected static List propagateDBRefsToCDS(SequenceI cdsSeq, SequenceI contig, SequenceI proteinProduct, Mapping mapping) { // gather direct refs from contig congruent with mapping List direct = new ArrayList<>(); HashSet directSources = new HashSet<>(); if (contig.getDBRefs() != null) { for (DBRefEntry dbr : contig.getDBRefs()) { if (dbr.hasMap() && dbr.getMap().getMap().isTripletMap()) { MapList map = dbr.getMap().getMap(); // check if map is the CDS mapping if (mapping.getMap().equals(map)) { direct.add(dbr); directSources.add(dbr.getSource()); } } } } DBRefEntry[] onSource = DBRefUtils.selectRefs( proteinProduct.getDBRefs(), directSources.toArray(new String[0])); List propagated = new ArrayList<>(); // and generate appropriate mappings for (DBRefEntry cdsref : direct) { // clone maplist and mapping MapList cdsposmap = new MapList( Arrays.asList(new int[][] { new int[] { cdsSeq.getStart(), cdsSeq.getEnd() } }), cdsref.getMap().getMap().getToRanges(), 3, 1); Mapping cdsmap = new Mapping(cdsref.getMap().getTo(), cdsref.getMap().getMap()); // create dbref DBRefEntry newref = new DBRefEntry(cdsref.getSource(), cdsref.getVersion(), cdsref.getAccessionId(), new Mapping(cdsmap.getTo(), cdsposmap)); // and see if we can map to the protein product for this mapping. // onSource is the filtered set of accessions on protein that we are // tranferring, so we assume accession is the same. if (cdsmap.getTo() == null && onSource != null) { List sourceRefs = DBRefUtils.searchRefs(onSource, cdsref.getAccessionId()); if (sourceRefs != null) { for (DBRefEntry srcref : sourceRefs) { if (srcref.getSource().equalsIgnoreCase(cdsref.getSource())) { // we have found a complementary dbref on the protein product, so // update mapping's getTo newref.getMap().setTo(proteinProduct); } } } } cdsSeq.addDBRef(newref); propagated.add(newref); } return propagated; } /** * Transfers co-located features on 'fromSeq' to 'toSeq', adjusting the * feature start/end ranges, optionally omitting specified feature types. * Returns the number of features copied. * * @param fromSeq * @param toSeq * @param mapping * the mapping from 'fromSeq' to 'toSeq' * @param select * if not null, only features of this type are copied (including * subtypes in the Sequence Ontology) * @param omitting */ protected static int transferFeatures(SequenceI fromSeq, SequenceI toSeq, MapList mapping, String select, String... omitting) { SequenceI copyTo = toSeq; while (copyTo.getDatasetSequence() != null) { copyTo = copyTo.getDatasetSequence(); } /* * get features, optionally restricted by an ontology term */ List sfs = select == null ? fromSeq.getFeatures() .getPositionalFeatures() : fromSeq.getFeatures() .getFeaturesByOntology(select); int count = 0; for (SequenceFeature sf : sfs) { String type = sf.getType(); boolean omit = false; for (String toOmit : omitting) { if (type.equals(toOmit)) { omit = true; } } if (omit) { continue; } /* * locate the mapped range - null if either start or end is * not mapped (no partial overlaps are calculated) */ int start = sf.getBegin(); int end = sf.getEnd(); int[] mappedTo = mapping.locateInTo(start, end); /* * if whole exon range doesn't map, try interpreting it * as 5' or 3' exon overlapping the CDS range */ if (mappedTo == null) { mappedTo = mapping.locateInTo(end, end); if (mappedTo != null) { /* * end of exon is in CDS range - 5' overlap * to a range from the start of the peptide */ mappedTo[0] = 1; } } if (mappedTo == null) { mappedTo = mapping.locateInTo(start, start); if (mappedTo != null) { /* * start of exon is in CDS range - 3' overlap * to a range up to the end of the peptide */ mappedTo[1] = toSeq.getLength(); } } if (mappedTo != null) { int newBegin = Math.min(mappedTo[0], mappedTo[1]); int newEnd = Math.max(mappedTo[0], mappedTo[1]); SequenceFeature copy = new SequenceFeature(sf, newBegin, newEnd, sf.getFeatureGroup(), sf.getScore()); copyTo.addSequenceFeature(copy); count++; } } return count; } /** * Returns a mapping from dna to protein by inspecting sequence features of * type "CDS" on the dna. A mapping is constructed if the total CDS feature * length is 3 times the peptide length (optionally after dropping a trailing * stop codon). This method does not check whether the CDS nucleotide sequence * translates to the peptide sequence. * * @param dnaSeq * @param proteinSeq * @return */ public static MapList mapCdsToProtein(SequenceI dnaSeq, SequenceI proteinSeq) { List ranges = findCdsPositions(dnaSeq); int mappedDnaLength = MappingUtils.getLength(ranges); /* * if not a whole number of codons, truncate mapping */ int codonRemainder = mappedDnaLength % CODON_LENGTH; if (codonRemainder > 0) { mappedDnaLength -= codonRemainder; MappingUtils.removeEndPositions(codonRemainder, ranges); } int proteinLength = proteinSeq.getLength(); int proteinStart = proteinSeq.getStart(); int proteinEnd = proteinSeq.getEnd(); /* * incomplete start codon may mean X at start of peptide * we ignore both for mapping purposes */ if (proteinSeq.getCharAt(0) == 'X') { // todo JAL-2022 support startPhase > 0 proteinStart++; proteinLength--; } List proteinRange = new ArrayList<>(); /* * dna length should map to protein (or protein plus stop codon) */ int codesForResidues = mappedDnaLength / CODON_LENGTH; if (codesForResidues == (proteinLength + 1)) { // assuming extra codon is for STOP and not in peptide // todo: check trailing codon is indeed a STOP codon codesForResidues--; mappedDnaLength -= CODON_LENGTH; MappingUtils.removeEndPositions(CODON_LENGTH, ranges); } if (codesForResidues == proteinLength) { proteinRange.add(new int[] { proteinStart, proteinEnd }); return new MapList(ranges, proteinRange, CODON_LENGTH, 1); } return null; } /** * Returns a list of CDS ranges found (as sequence positions base 1), i.e. of * [start, end] positions of sequence features of type "CDS" (or a sub-type of * CDS in the Sequence Ontology). The ranges are sorted into ascending start * position order, so this method is only valid for linear CDS in the same * sense as the protein product. * * @param dnaSeq * @return */ protected static List findCdsPositions(SequenceI dnaSeq) { List result = new ArrayList<>(); List sfs = dnaSeq.getFeatures().getFeaturesByOntology( SequenceOntologyI.CDS); if (sfs.isEmpty()) { return result; } SequenceFeatures.sortFeatures(sfs, true); for (SequenceFeature sf : sfs) { int phase = 0; try { phase = Integer.parseInt(sf.getPhase()); } catch (NumberFormatException e) { // ignore } /* * phase > 0 on first codon means 5' incomplete - skip to the start * of the next codon; example ENST00000496384 */ int begin = sf.getBegin(); int end = sf.getEnd(); if (result.isEmpty() && phase > 0) { begin += phase; if (begin > end) { // shouldn't happen! System.err .println("Error: start phase extends beyond start CDS in " + dnaSeq.getName()); } } result.add(new int[] { begin, end }); } /* * Finally sort ranges by start position. This avoids a dependency on * keeping features in order on the sequence (if they are in order anyway, * the sort will have almost no work to do). The implicit assumption is CDS * ranges are assembled in order. Other cases should not use this method, * but instead construct an explicit mapping for CDS (e.g. EMBL parsing). */ Collections.sort(result, IntRangeComparator.ASCENDING); return result; } /** * Maps exon features from dna to protein, and computes variants in peptide * product generated by variants in dna, and adds them as sequence_variant * features on the protein sequence. Returns the number of variant features * added. * * @param dnaSeq * @param peptide * @param dnaToProtein */ public static int computeProteinFeatures(SequenceI dnaSeq, SequenceI peptide, MapList dnaToProtein) { while (dnaSeq.getDatasetSequence() != null) { dnaSeq = dnaSeq.getDatasetSequence(); } while (peptide.getDatasetSequence() != null) { peptide = peptide.getDatasetSequence(); } transferFeatures(dnaSeq, peptide, dnaToProtein, SequenceOntologyI.EXON); /* * compute protein variants from dna variants and codon mappings; * NB - alternatively we could retrieve this using the REST service e.g. * http://rest.ensembl.org/overlap/translation * /ENSP00000288602?feature=transcript_variation;content-type=text/xml * which would be a bit slower but possibly more reliable */ /* * build a map with codon variations for each potentially varying peptide */ LinkedHashMap[]> variants = buildDnaVariantsMap( dnaSeq, dnaToProtein); /* * scan codon variations, compute peptide variants and add to peptide sequence */ int count = 0; for (Entry[]> variant : variants.entrySet()) { int peptidePos = variant.getKey(); List[] codonVariants = variant.getValue(); count += computePeptideVariants(peptide, peptidePos, codonVariants); } return count; } /** * Computes non-synonymous peptide variants from codon variants and adds them * as sequence_variant features on the protein sequence (one feature per * allele variant). Selected attributes (variant id, clinical significance) * are copied over to the new features. * * @param peptide * the protein sequence * @param peptidePos * the position to compute peptide variants for * @param codonVariants * a list of dna variants per codon position * @return the number of features added */ static int computePeptideVariants(SequenceI peptide, int peptidePos, List[] codonVariants) { String residue = String.valueOf(peptide.getCharAt(peptidePos - 1)); int count = 0; String base1 = codonVariants[0].get(0).base; String base2 = codonVariants[1].get(0).base; String base3 = codonVariants[2].get(0).base; /* * variants in first codon base */ for (DnaVariant var : codonVariants[0]) { if (var.variant != null) { String alleles = (String) var.variant.getValue(Gff3Helper.ALLELES); if (alleles != null) { for (String base : alleles.split(",")) { if (!base1.equalsIgnoreCase(base)) { String codon = base.toUpperCase() + base2.toLowerCase() + base3.toLowerCase(); String canonical = base1.toUpperCase() + base2.toLowerCase() + base3.toLowerCase(); if (addPeptideVariant(peptide, peptidePos, residue, var, codon, canonical)) { count++; } } } } } } /* * variants in second codon base */ for (DnaVariant var : codonVariants[1]) { if (var.variant != null) { String alleles = (String) var.variant.getValue(Gff3Helper.ALLELES); if (alleles != null) { for (String base : alleles.split(",")) { if (!base2.equalsIgnoreCase(base)) { String codon = base1.toLowerCase() + base.toUpperCase() + base3.toLowerCase(); String canonical = base1.toLowerCase() + base2.toUpperCase() + base3.toLowerCase(); if (addPeptideVariant(peptide, peptidePos, residue, var, codon, canonical)) { count++; } } } } } } /* * variants in third codon base */ for (DnaVariant var : codonVariants[2]) { if (var.variant != null) { String alleles = (String) var.variant.getValue(Gff3Helper.ALLELES); if (alleles != null) { for (String base : alleles.split(",")) { if (!base3.equalsIgnoreCase(base)) { String codon = base1.toLowerCase() + base2.toLowerCase() + base.toUpperCase(); String canonical = base1.toLowerCase() + base2.toLowerCase() + base3.toUpperCase(); if (addPeptideVariant(peptide, peptidePos, residue, var, codon, canonical)) { count++; } } } } } } return count; } /** * Helper method that adds a peptide variant feature. ID and * clinical_significance attributes of the dna variant (if present) are copied * to the new feature. * * @param peptide * @param peptidePos * @param residue * @param var * @param codon * the variant codon e.g. aCg * @param canonical * the 'normal' codon e.g. aTg * @return true if a feature was added, else false */ static boolean addPeptideVariant(SequenceI peptide, int peptidePos, String residue, DnaVariant var, String codon, String canonical) { /* * get peptide translation of codon e.g. GAT -> D * note that variants which are not single alleles, * e.g. multibase variants or HGMD_MUTATION etc * are currently ignored here */ String trans = codon.contains("-") ? null : (codon.length() > CODON_LENGTH ? null : ResidueProperties.codonTranslate(codon)); if (trans == null) { return false; } String desc = canonical + "/" + codon; String featureType = ""; if (trans.equals(residue)) { featureType = SequenceOntologyI.SYNONYMOUS_VARIANT; } else if (ResidueProperties.STOP.equals(trans)) { featureType = SequenceOntologyI.STOP_GAINED; } else { String residue3Char = StringUtils .toSentenceCase(ResidueProperties.aa2Triplet.get(residue)); String trans3Char = StringUtils .toSentenceCase(ResidueProperties.aa2Triplet.get(trans)); desc = "p." + residue3Char + peptidePos + trans3Char; featureType = SequenceOntologyI.NONSYNONYMOUS_VARIANT; } SequenceFeature sf = new SequenceFeature(featureType, desc, peptidePos, peptidePos, var.getSource()); StringBuilder attributes = new StringBuilder(32); String id = (String) var.variant.getValue(VARIANT_ID); if (id != null) { if (id.startsWith(SEQUENCE_VARIANT)) { id = id.substring(SEQUENCE_VARIANT.length()); } sf.setValue(VARIANT_ID, id); attributes.append(VARIANT_ID).append("=").append(id); // TODO handle other species variants JAL-2064 StringBuilder link = new StringBuilder(32); try { link.append(desc).append(" ").append(id).append( "|http://www.ensembl.org/Homo_sapiens/Variation/Summary?v=") .append(URLEncoder.encode(id, "UTF-8")); sf.addLink(link.toString()); } catch (UnsupportedEncodingException e) { // as if } } String clinSig = (String) var.variant.getValue(CLINICAL_SIGNIFICANCE); if (clinSig != null) { sf.setValue(CLINICAL_SIGNIFICANCE, clinSig); attributes.append(";").append(CLINICAL_SIGNIFICANCE).append("=") .append(clinSig); } peptide.addSequenceFeature(sf); if (attributes.length() > 0) { sf.setAttributes(attributes.toString()); } return true; } /** * Builds a map whose key is position in the protein sequence, and value is a * list of the base and all variants for each corresponding codon position. *

* This depends on dna variants being held as a comma-separated list as * property "alleles" on variant features. * * @param dnaSeq * @param dnaToProtein * @return */ @SuppressWarnings("unchecked") static LinkedHashMap[]> buildDnaVariantsMap( SequenceI dnaSeq, MapList dnaToProtein) { /* * map from peptide position to all variants of the codon which codes for it * LinkedHashMap ensures we keep the peptide features in sequence order */ LinkedHashMap[]> variants = new LinkedHashMap<>(); List dnaFeatures = dnaSeq.getFeatures() .getFeaturesByOntology(SequenceOntologyI.SEQUENCE_VARIANT); if (dnaFeatures.isEmpty()) { return variants; } int dnaStart = dnaSeq.getStart(); int[] lastCodon = null; int lastPeptidePostion = 0; /* * build a map of codon variations for peptides */ for (SequenceFeature sf : dnaFeatures) { int dnaCol = sf.getBegin(); if (dnaCol != sf.getEnd()) { // not handling multi-locus variant features continue; } /* * ignore variant if not a SNP */ String alls = (String) sf.getValue(Gff3Helper.ALLELES); if (alls == null) { continue; // non-SNP VCF variant perhaps - can't process this } String[] alleles = alls.toUpperCase().split(","); boolean isSnp = true; for (String allele : alleles) { if (allele.trim().length() > 1) { isSnp = false; } } if (!isSnp) { continue; } int[] mapsTo = dnaToProtein.locateInTo(dnaCol, dnaCol); if (mapsTo == null) { // feature doesn't lie within coding region continue; } int peptidePosition = mapsTo[0]; List[] codonVariants = variants.get(peptidePosition); if (codonVariants == null) { codonVariants = new ArrayList[CODON_LENGTH]; codonVariants[0] = new ArrayList<>(); codonVariants[1] = new ArrayList<>(); codonVariants[2] = new ArrayList<>(); variants.put(peptidePosition, codonVariants); } /* * get this peptide's codon positions e.g. [3, 4, 5] or [4, 7, 10] */ int[] codon = peptidePosition == lastPeptidePostion ? lastCodon : MappingUtils.flattenRanges(dnaToProtein.locateInFrom( peptidePosition, peptidePosition)); lastPeptidePostion = peptidePosition; lastCodon = codon; /* * save nucleotide (and any variant) for each codon position */ for (int codonPos = 0; codonPos < CODON_LENGTH; codonPos++) { String nucleotide = String.valueOf( dnaSeq.getCharAt(codon[codonPos] - dnaStart)).toUpperCase(); List codonVariant = codonVariants[codonPos]; if (codon[codonPos] == dnaCol) { if (!codonVariant.isEmpty() && codonVariant.get(0).variant == null) { /* * already recorded base value, add this variant */ codonVariant.get(0).variant = sf; } else { /* * add variant with base value */ codonVariant.add(new DnaVariant(nucleotide, sf)); } } else if (codonVariant.isEmpty()) { /* * record (possibly non-varying) base value */ codonVariant.add(new DnaVariant(nucleotide)); } } } return variants; } /** * Makes an alignment with a copy of the given sequences, adding in any * non-redundant sequences which are mapped to by the cross-referenced * sequences. * * @param seqs * @param xrefs * @param dataset * the alignment dataset shared by the new copy * @return */ public static AlignmentI makeCopyAlignment(SequenceI[] seqs, SequenceI[] xrefs, AlignmentI dataset) { AlignmentI copy = new Alignment(new Alignment(seqs)); copy.setDataset(dataset); boolean isProtein = !copy.isNucleotide(); SequenceIdMatcher matcher = new SequenceIdMatcher(seqs); if (xrefs != null) { for (SequenceI xref : xrefs) { DBRefEntry[] dbrefs = xref.getDBRefs(); if (dbrefs != null) { for (DBRefEntry dbref : dbrefs) { if (dbref.getMap() == null || dbref.getMap().getTo() == null || dbref.getMap().getTo().isProtein() != isProtein) { continue; } SequenceI mappedTo = dbref.getMap().getTo(); SequenceI match = matcher.findIdMatch(mappedTo); if (match == null) { matcher.add(mappedTo); copy.addSequence(mappedTo); } } } } } return copy; } /** * Try to align sequences in 'unaligned' to match the alignment of their * mapped regions in 'aligned'. For example, could use this to align CDS * sequences which are mapped to their parent cDNA sequences. * * This method handles 1:1 mappings (dna-to-dna or protein-to-protein). For * dna-to-protein or protein-to-dna use alternative methods. * * @param unaligned * sequences to be aligned * @param aligned * holds aligned sequences and their mappings * @return */ public static int alignAs(AlignmentI unaligned, AlignmentI aligned) { /* * easy case - aligning a copy of aligned sequences */ if (alignAsSameSequences(unaligned, aligned)) { return unaligned.getHeight(); } /* * fancy case - aligning via mappings between sequences */ List unmapped = new ArrayList<>(); Map> columnMap = buildMappedColumnsMap( unaligned, aligned, unmapped); int width = columnMap.size(); char gap = unaligned.getGapCharacter(); int realignedCount = 0; // TODO: verify this loop scales sensibly for very wide/high alignments for (SequenceI seq : unaligned.getSequences()) { if (!unmapped.contains(seq)) { char[] newSeq = new char[width]; Arrays.fill(newSeq, gap); // JBPComment - doubt this is faster than the // Integer iteration below int newCol = 0; int lastCol = 0; /* * traverse the map to find columns populated * by our sequence */ for (Integer column : columnMap.keySet()) { Character c = columnMap.get(column).get(seq); if (c != null) { /* * sequence has a character at this position * */ newSeq[newCol] = c; lastCol = newCol; } newCol++; } /* * trim trailing gaps */ if (lastCol < width) { char[] tmp = new char[lastCol + 1]; System.arraycopy(newSeq, 0, tmp, 0, lastCol + 1); newSeq = tmp; } // TODO: optimise SequenceI to avoid char[]->String->char[] seq.setSequence(String.valueOf(newSeq)); realignedCount++; } } return realignedCount; } /** * If unaligned and aligned sequences share the same dataset sequences, then * simply copies the aligned sequences to the unaligned sequences and returns * true; else returns false * * @param unaligned * - sequences to be aligned based on aligned * @param aligned * - 'guide' alignment containing sequences derived from same dataset * as unaligned * @return */ static boolean alignAsSameSequences(AlignmentI unaligned, AlignmentI aligned) { if (aligned.getDataset() == null || unaligned.getDataset() == null) { return false; // should only pass alignments with datasets here } // map from dataset sequence to alignment sequence(s) Map> alignedDatasets = new HashMap<>(); for (SequenceI seq : aligned.getSequences()) { SequenceI ds = seq.getDatasetSequence(); if (alignedDatasets.get(ds) == null) { alignedDatasets.put(ds, new ArrayList()); } alignedDatasets.get(ds).add(seq); } /* * first pass - check whether all sequences to be aligned share a dataset * sequence with an aligned sequence */ for (SequenceI seq : unaligned.getSequences()) { if (!alignedDatasets.containsKey(seq.getDatasetSequence())) { return false; } } /* * second pass - copy aligned sequences; * heuristic rule: pair off sequences in order for the case where * more than one shares the same dataset sequence */ for (SequenceI seq : unaligned.getSequences()) { List alignedSequences = alignedDatasets .get(seq.getDatasetSequence()); // TODO: getSequenceAsString() will be deprecated in the future // TODO: need to leave to SequenceI implementor to update gaps seq.setSequence(alignedSequences.get(0).getSequenceAsString()); if (alignedSequences.size() > 0) { // pop off aligned sequences (except the last one) alignedSequences.remove(0); } } return true; } /** * Returns a map whose key is alignment column number (base 1), and whose * values are a map of sequence characters in that column. * * @param unaligned * @param aligned * @param unmapped * @return */ static SortedMap> buildMappedColumnsMap( AlignmentI unaligned, AlignmentI aligned, List unmapped) { /* * Map will hold, for each aligned column position, a map of * {unalignedSequence, characterPerSequence} at that position. * TreeMap keeps the entries in ascending column order. */ SortedMap> map = new TreeMap<>(); /* * record any sequences that have no mapping so can't be realigned */ unmapped.addAll(unaligned.getSequences()); List mappings = aligned.getCodonFrames(); for (SequenceI seq : unaligned.getSequences()) { for (AlignedCodonFrame mapping : mappings) { SequenceI fromSeq = mapping.findAlignedSequence(seq, aligned); if (fromSeq != null) { Mapping seqMap = mapping.getMappingBetween(fromSeq, seq); if (addMappedPositions(seq, fromSeq, seqMap, map)) { unmapped.remove(seq); } } } } return map; } /** * Helper method that adds to a map the mapped column positions of a sequence. *
* For example if aaTT-Tg-gAAA is mapped to TTTAAA then the map should record * that columns 3,4,6,10,11,12 map to characters T,T,T,A,A,A of the mapped to * sequence. * * @param seq * the sequence whose column positions we are recording * @param fromSeq * a sequence that is mapped to the first sequence * @param seqMap * the mapping from 'fromSeq' to 'seq' * @param map * a map to add the column positions (in fromSeq) of the mapped * positions of seq * @return */ static boolean addMappedPositions(SequenceI seq, SequenceI fromSeq, Mapping seqMap, Map> map) { if (seqMap == null) { return false; } /* * invert mapping if it is from unaligned to aligned sequence */ if (seqMap.getTo() == fromSeq.getDatasetSequence()) { seqMap = new Mapping(seq.getDatasetSequence(), seqMap.getMap().getInverse()); } int toStart = seq.getStart(); /* * traverse [start, end, start, end...] ranges in fromSeq */ for (int[] fromRange : seqMap.getMap().getFromRanges()) { for (int i = 0; i < fromRange.length - 1; i += 2) { boolean forward = fromRange[i + 1] >= fromRange[i]; /* * find the range mapped to (sequence positions base 1) */ int[] range = seqMap.locateMappedRange(fromRange[i], fromRange[i + 1]); if (range == null) { System.err.println("Error in mapping " + seqMap + " from " + fromSeq.getName()); return false; } int fromCol = fromSeq.findIndex(fromRange[i]); int mappedCharPos = range[0]; /* * walk over the 'from' aligned sequence in forward or reverse * direction; when a non-gap is found, record the column position * of the next character of the mapped-to sequence; stop when all * the characters of the range have been counted */ while (mappedCharPos <= range[1] && fromCol <= fromSeq.getLength() && fromCol >= 0) { if (!Comparison.isGap(fromSeq.getCharAt(fromCol - 1))) { /* * mapped from sequence has a character in this column * record the column position for the mapped to character */ Map seqsMap = map.get(fromCol); if (seqsMap == null) { seqsMap = new HashMap<>(); map.put(fromCol, seqsMap); } seqsMap.put(seq, seq.getCharAt(mappedCharPos - toStart)); mappedCharPos++; } fromCol += (forward ? 1 : -1); } } } return true; } // strictly temporary hack until proper criteria for aligning protein to cds // are in place; this is so Ensembl -> fetch xrefs Uniprot aligns the Uniprot public static boolean looksLikeEnsembl(AlignmentI alignment) { for (SequenceI seq : alignment.getSequences()) { String name = seq.getName(); if (!name.startsWith("ENSG") && !name.startsWith("ENST")) { return false; } } return true; } }