+ /*
+ * 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|<dbref>
+ // <dbref> 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<int[]> 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();
+ MapList dnaToCdsMap = new MapList(mapList.getFromRanges(),
+ cdsRange, 1, 1);
+ dnaToCdsMapping.addMap(dnaSeq.getDatasetSequence(), cdsSeqDss,
+ dnaToCdsMap);
+ if (!mappings.contains(dnaToCdsMapping))
+ {
+ mappings.add(dnaToCdsMapping);
+ }
+
+ /*
+ * 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())
+ {
+ // creates a complementary cross-reference to the source sequence's
+ // primary reference.
+
+ DBRefEntry cdsCrossRef = new DBRefEntry(primRef.getSource(),
+ primRef.getSource() + ":" + primRef.getVersion(),
+ primRef.getAccessionId());
+ cdsCrossRef
+ .setMap(new Mapping(dnaDss, new MapList(dnaToCdsMap)));
+ cdsSeqDss.addDBRef(cdsCrossRef);
+
+ // 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(primRef.getSource(),
+ primRef.getVersion(), 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;
+ }
+
+ /**
+ * 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
+ * an initial candidate from seqMappings
+ * @return
+ */
+ static SequenceI findCdsForProtein(List<AlignedCodonFrame> mappings,
+ SequenceI dnaSeq, List<AlignedCodonFrame> 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)
+ {
+ return seqDss;
+ }
+
+ /*
+ * looks like we found the dna-to-protein mapping; search for the
+ * corresponding cds-to-protein mapping
+ */
+ List<AlignedCodonFrame> 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 it
+ * is from the dna start sequence
+ */
+ SequenceI cdsSeq = map.getFromSeq();
+ List<AlignedCodonFrame> 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<int[]> 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;
+ }
+
+ /**
+ * add any DBRefEntrys to cdsSeq from contig that have a Mapping congruent to
+ * the given mapping.
+ *
+ * @param cdsSeq
+ * @param contig
+ * @param mapping
+ * @return list of DBRefEntrys added.
+ */
+ public static List<DBRefEntry> propagateDBRefsToCDS(SequenceI cdsSeq,
+ SequenceI contig, SequenceI proteinProduct, Mapping mapping)
+ {
+
+ // gather direct refs from contig congrent with mapping
+ List<DBRefEntry> direct = new ArrayList<DBRefEntry>();
+ HashSet<String> directSources = new HashSet<String>();
+ 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<DBRefEntry> propagated = new ArrayList<DBRefEntry>();
+
+ // 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<DBRefEntry> 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 select
+ * if not null, only features of this type are copied (including
+ * subtypes in the Sequence Ontology)
+ * @param mapping
+ * the mapping from 'fromSeq' to 'toSeq'
+ * @param omitting
+ */
+ public static int transferFeatures(SequenceI fromSeq, SequenceI toSeq,
+ MapList mapping, String select, String... omitting)
+ {
+ SequenceI copyTo = toSeq;
+ while (copyTo.getDatasetSequence() != null)
+ {
+ copyTo = copyTo.getDatasetSequence();
+ }
+
+ SequenceOntologyI so = SequenceOntologyFactory.getInstance();
+ int count = 0;
+ SequenceFeature[] sfs = fromSeq.getSequenceFeatures();
+ if (sfs != null)
+ {
+ for (SequenceFeature sf : sfs)
+ {
+ String type = sf.getType();
+ if (select != null && !so.isA(type, select))
+ {
+ continue;
+ }
+ 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)
+ {
+ SequenceFeature copy = new SequenceFeature(sf);
+ copy.setBegin(Math.min(mappedTo[0], mappedTo[1]));
+ copy.setEnd(Math.max(mappedTo[0], mappedTo[1]));
+ copyTo.addSequenceFeature(copy);
+ count++;
+ }
+ }
+ }
+ return count;
+ }
+
+ /**
+ * Returns a mapping from dna to protein by inspecting sequence features of
+ * type "CDS" on the dna.
+ *
+ * @param dnaSeq
+ * @param proteinSeq
+ * @return
+ */
+ public static MapList mapCdsToProtein(SequenceI dnaSeq,
+ SequenceI proteinSeq)
+ {
+ List<int[]> ranges = findCdsPositions(dnaSeq);
+ int mappedDnaLength = MappingUtils.getLength(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<int[]> proteinRange = new ArrayList<int[]>();
+
+ /*
+ * 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
+ codesForResidues--;
+ }
+ 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
+ */
+ public static List<int[]> findCdsPositions(SequenceI dnaSeq)
+ {
+ List<int[]> result = new ArrayList<int[]>();
+ SequenceFeature[] sfs = dnaSeq.getSequenceFeatures();
+ if (sfs == null)
+ {
+ return result;
+ }
+
+ SequenceOntologyI so = SequenceOntologyFactory.getInstance();
+ int startPhase = 0;
+
+ for (SequenceFeature sf : sfs)
+ {
+ /*
+ * process a CDS feature (or a sub-type of CDS)
+ */
+ if (so.isA(sf.getType(), SequenceOntologyI.CDS))
+ {
+ 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())
+ {
+ 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 });
+ }
+ }
+
+ /*
+ * remove 'startPhase' positions (usually 0) from the first range
+ * so we begin at the start of a complete codon
+ */
+ if (!result.isEmpty())
+ {
+ // TODO JAL-2022 correctly model start phase > 0
+ result.get(0)[0] += startPhase;
+ }
+
+ /*
+ * 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, new RangeComparator(true));
+ 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<Integer, List<DnaVariant>[]> variants = buildDnaVariantsMap(
+ dnaSeq, dnaToProtein);
+
+ /*
+ * scan codon variations, compute peptide variants and add to peptide sequence
+ */
+ int count = 0;
+ for (Entry<Integer, List<DnaVariant>[]> variant : variants.entrySet())
+ {
+ int peptidePos = variant.getKey();
+ List<DnaVariant>[] codonVariants = variant.getValue();
+ count += computePeptideVariants(peptide, peptidePos, codonVariants);
+ }
+
+ /*
+ * sort to get sequence features in start position order
+ * - would be better to store in Sequence as a TreeSet or NCList?
+ */
+ if (peptide.getSequenceFeatures() != null)
+ {
+ Arrays.sort(peptide.getSequenceFeatures(),
+ new Comparator<SequenceFeature>()
+ {
+ @Override
+ public int compare(SequenceFeature o1, SequenceFeature o2)
+ {
+ int c = Integer.compare(o1.getBegin(), o2.getBegin());
+ return c == 0 ? Integer.compare(o1.getEnd(), o2.getEnd())
+ : c;
+ }
+ });
+ }
+ 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<DnaVariant>[] 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("alleles");
+ if (alleles != null)
+ {
+ for (String base : alleles.split(","))
+ {
+ String codon = base + base2 + base3;
+ if (addPeptideVariant(peptide, peptidePos, residue, var, codon))
+ {
+ count++;
+ }
+ }
+ }
+ }
+ }
+
+ /*
+ * variants in second codon base
+ */
+ for (DnaVariant var : codonVariants[1])
+ {
+ if (var.variant != null)
+ {
+ String alleles = (String) var.variant.getValue("alleles");
+ if (alleles != null)
+ {
+ for (String base : alleles.split(","))
+ {
+ String codon = base1 + base + base3;
+ if (addPeptideVariant(peptide, peptidePos, residue, var, codon))
+ {
+ count++;
+ }
+ }
+ }
+ }
+ }
+
+ /*
+ * variants in third codon base
+ */
+ for (DnaVariant var : codonVariants[2])
+ {
+ if (var.variant != null)
+ {
+ String alleles = (String) var.variant.getValue("alleles");
+ if (alleles != null)
+ {
+ for (String base : alleles.split(","))
+ {
+ String codon = base1 + base2 + base;
+ if (addPeptideVariant(peptide, peptidePos, residue, var, codon))
+ {
+ count++;
+ }
+ }
+ }
+ }
+ }
+
+ return count;
+ }
+
+ /**
+ * Helper method that adds a peptide variant feature, provided the given codon
+ * translates to a value different to the current residue (is a non-synonymous
+ * variant). 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
+ * @return true if a feature was added, else false
+ */
+ static boolean addPeptideVariant(SequenceI peptide, int peptidePos,
+ String residue, DnaVariant var, String codon)
+ {
+ /*
+ * 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("-") ? "-"
+ : (codon.length() > CODON_LENGTH ? null
+ : ResidueProperties.codonTranslate(codon));
+ if (trans != null && !trans.equals(residue))
+ {
+ String residue3Char = StringUtils
+ .toSentenceCase(ResidueProperties.aa2Triplet.get(residue));
+ String trans3Char = StringUtils
+ .toSentenceCase(ResidueProperties.aa2Triplet.get(trans));
+ String desc = "p." + residue3Char + peptidePos + trans3Char;
+ // set score to 0f so 'graduated colour' option is offered! JAL-2060
+ SequenceFeature sf = new SequenceFeature(
+ SequenceOntologyI.SEQUENCE_VARIANT, desc, peptidePos,
+ peptidePos, 0f, var.getSource());
+ StringBuilder attributes = new StringBuilder(32);
+ String id = (String) var.variant.getValue(ID);
+ if (id != null)
+ {
+ if (id.startsWith(SEQUENCE_VARIANT))
+ {
+ id = id.substring(SEQUENCE_VARIANT.length());
+ }
+ sf.setValue(ID, id);
+ attributes.append(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;
+ }
+ return false;
+ }
+
+ /**
+ * 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
+ *
+ * @param dnaSeq
+ * @param dnaToProtein
+ * @return
+ */
+ @SuppressWarnings("unchecked")
+ static LinkedHashMap<Integer, List<DnaVariant>[]> 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<Integer, List<DnaVariant>[]> variants = new LinkedHashMap<Integer, List<DnaVariant>[]>();
+ SequenceOntologyI so = SequenceOntologyFactory.getInstance();
+
+ SequenceFeature[] dnaFeatures = dnaSeq.getSequenceFeatures();
+ if (dnaFeatures == null)
+ {
+ 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;
+ }
+ if (so.isA(sf.getType(), SequenceOntologyI.SEQUENCE_VARIANT))
+ {
+ int[] mapsTo = dnaToProtein.locateInTo(dnaCol, dnaCol);
+ if (mapsTo == null)
+ {
+ // feature doesn't lie within coding region
+ continue;
+ }
+ int peptidePosition = mapsTo[0];
+ List<DnaVariant>[] codonVariants = variants.get(peptidePosition);
+ if (codonVariants == null)
+ {
+ codonVariants = new ArrayList[CODON_LENGTH];
+ codonVariants[0] = new ArrayList<DnaVariant>();
+ codonVariants[1] = new ArrayList<DnaVariant>();
+ codonVariants[2] = new ArrayList<DnaVariant>();
+ variants.put(peptidePosition, codonVariants);
+ }
+
+ /*
+ * extract dna variants to a string array
+ */
+ String alls = (String) sf.getValue("alleles");
+ if (alls == null)
+ {
+ continue;
+ }
+ String[] alleles = alls.toUpperCase().split(",");
+ int i = 0;
+ for (String allele : alleles)
+ {
+ alleles[i++] = allele.trim(); // lose any space characters "A, G"
+ }
+
+ /*
+ * 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<DnaVariant> 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<SequenceI> unmapped = new ArrayList<SequenceI>();
+ Map<Integer, Map<SequenceI, Character>> 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<SequenceI, List<SequenceI>> alignedDatasets = new HashMap<SequenceI, List<SequenceI>>();
+ for (SequenceI seq : aligned.getSequences())
+ {
+ SequenceI ds = seq.getDatasetSequence();
+ if (alignedDatasets.get(ds) == null)
+ {
+ alignedDatasets.put(ds, new ArrayList<SequenceI>());
+ }
+ 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<SequenceI> 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<Integer, Map<SequenceI, Character>> buildMappedColumnsMap(
+ AlignmentI unaligned, AlignmentI aligned,
+ List<SequenceI> 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<Integer, Map<SequenceI, Character>> map = new TreeMap<Integer, Map<SequenceI, Character>>();
+
+ /*
+ * record any sequences that have no mapping so can't be realigned
+ */
+ unmapped.addAll(unaligned.getSequences());
+
+ List<AlignedCodonFrame> 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.
+ * <br>
+ * 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<Integer, Map<SequenceI, Character>> 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());
+ }
+
+ char[] fromChars = fromSeq.getSequence();
+ int toStart = seq.getStart();
+ char[] toChars = seq.getSequence();