1 package jalview.io.vcf;
3 import htsjdk.samtools.util.CloseableIterator;
4 import htsjdk.variant.variantcontext.Allele;
5 import htsjdk.variant.variantcontext.VariantContext;
6 import htsjdk.variant.vcf.VCFHeader;
7 import htsjdk.variant.vcf.VCFHeaderLine;
8 import htsjdk.variant.vcf.VCFHeaderLineCount;
9 import htsjdk.variant.vcf.VCFInfoHeaderLine;
11 import jalview.analysis.AlignmentUtils;
12 import jalview.analysis.Dna;
13 import jalview.api.AlignViewControllerGuiI;
14 import jalview.datamodel.AlignmentI;
15 import jalview.datamodel.DBRefEntry;
16 import jalview.datamodel.GeneLociI;
17 import jalview.datamodel.Mapping;
18 import jalview.datamodel.SequenceFeature;
19 import jalview.datamodel.SequenceI;
20 import jalview.ext.ensembl.EnsemblMap;
21 import jalview.ext.htsjdk.VCFReader;
22 import jalview.io.gff.Gff3Helper;
23 import jalview.io.gff.SequenceOntologyI;
24 import jalview.util.MapList;
25 import jalview.util.MappingUtils;
26 import jalview.util.MessageManager;
28 import java.io.IOException;
29 import java.util.ArrayList;
30 import java.util.HashMap;
31 import java.util.List;
33 import java.util.Map.Entry;
36 * A class to read VCF data (using the htsjdk) and add variants as sequence
37 * features on dna and any related protein product sequences
41 public class VCFLoader
44 * keys to fields of VEP CSQ consequence data
45 * see https://www.ensembl.org/info/docs/tools/vep/vep_formats.html
47 private static final String ALLELE_KEY = "Allele";
49 private static final String ALLELE_NUM_KEY = "ALLELE_NUM"; // 0 (ref), 1...
50 private static final String FEATURE_KEY = "Feature"; // Ensembl stable id
53 * what comes before column headings in CSQ Description field
55 private static final String FORMAT = "Format: ";
58 * default VCF INFO key for VEP consequence data
59 * NB this can be overridden running VEP with --vcf_info_field
60 * - we don't handle this case (require CSQ identifier)
62 private static final String CSQ = "CSQ";
65 * separator for fields in consequence data
67 private static final String PIPE = "|";
69 private static final String PIPE_REGEX = "\\" + PIPE;
72 * key for Allele Frequency output by VEP
73 * see http://www.ensembl.org/info/docs/tools/vep/vep_formats.html
75 private static final String ALLELE_FREQUENCY_KEY = "AF";
78 * delimiter that separates multiple consequence data blocks
80 private static final String COMMA = ",";
83 * the feature group assigned to a VCF variant in Jalview
85 private static final String FEATURE_GROUP_VCF = "VCF";
88 * internal delimiter used to build keys for assemblyMappings
91 private static final String EXCL = "!";
94 * the alignment we are associating VCF data with
96 private AlignmentI al;
99 * mappings between VCF and sequence reference assembly regions, as
100 * key = "species!chromosome!fromAssembly!toAssembly
101 * value = Map{fromRange, toRange}
103 private Map<String, Map<int[], int[]>> assemblyMappings;
106 * holds details of the VCF header lines (metadata)
108 private VCFHeader header;
111 * the position (0...) of field in each block of
112 * CSQ (consequence) data (if declared in the VCF INFO header for CSQ)
113 * see http://www.ensembl.org/info/docs/tools/vep/vep_formats.html
115 private int csqAlleleFieldIndex = -1;
116 private int csqAlleleNumberFieldIndex = -1;
117 private int csqFeatureFieldIndex = -1;
120 * Constructor given an alignment context
124 public VCFLoader(AlignmentI alignment)
128 // map of species!chromosome!fromAssembly!toAssembly to {fromRange, toRange}
129 assemblyMappings = new HashMap<String, Map<int[], int[]>>();
133 * Starts a new thread to query and load VCF variant data on to the alignment
135 * This method is not thread safe - concurrent threads should use separate
136 * instances of this class.
141 public void loadVCF(final String filePath,
142 final AlignViewControllerGuiI gui)
146 gui.setStatus(MessageManager.getString("label.searching_vcf"));
155 VCFLoader.this.doLoad(filePath, gui);
162 * Loads VCF on to an alignment - provided it can be related to one or more
163 * sequence's chromosomal coordinates
167 * optional callback handler for messages
169 protected void doLoad(String filePath, AlignViewControllerGuiI gui)
171 VCFReader reader = null;
174 // long start = System.currentTimeMillis();
175 reader = new VCFReader(filePath);
177 header = reader.getFileHeader();
178 VCFHeaderLine ref = header
179 .getOtherHeaderLine(VCFHeader.REFERENCE_KEY);
182 * get offset of CSQ ALLELE_NUM and Feature if declared
186 // check if reference is wrt assembly19 (GRCh37)
187 // todo may need to allow user to specify reference assembly?
188 boolean isRefGrch37 = (ref != null && ref.getValue().contains(
195 * query for VCF overlapping each sequence in turn
197 for (SequenceI seq : al.getSequences())
199 int added = loadSequenceVCF(seq, reader, isRefGrch37);
204 transferAddedFeatures(seq);
209 // long elapsed = System.currentTimeMillis() - start;
210 String msg = MessageManager.formatMessage("label.added_vcf",
213 if (gui.getFeatureSettingsUI() != null)
215 gui.getFeatureSettingsUI().discoverAllFeatureData();
218 } catch (Throwable e)
220 System.err.println("Error processing VCF: " + e.getMessage());
224 gui.setStatus("Error occurred - see console for details");
233 } catch (IOException e)
242 * Records the position of selected fields defined in the CSQ INFO header (if
243 * there is one). CSQ fields are declared in the CSQ INFO Description e.g.
245 * Description="Consequence ...from ... VEP. Format: Allele|Consequence|...
247 protected void locateCsqFields()
249 VCFInfoHeaderLine csqInfo = header.getInfoHeaderLine(CSQ);
255 String desc = csqInfo.getDescription();
256 int formatPos = desc.indexOf(FORMAT);
259 System.err.println("Parse error, failed to find " + FORMAT
263 desc = desc.substring(formatPos + FORMAT.length());
267 String[] format = desc.split(PIPE_REGEX);
269 for (String field : format)
271 if (ALLELE_NUM_KEY.equals(field))
273 csqAlleleNumberFieldIndex = index;
275 if (ALLELE_KEY.equals(field))
277 csqAlleleFieldIndex = index;
279 if (FEATURE_KEY.equals(field))
281 csqFeatureFieldIndex = index;
289 * Transfers VCF features to sequences to which this sequence has a mapping.
290 * If the mapping is 1:3, computes peptide variants from nucleotide variants.
294 protected void transferAddedFeatures(SequenceI seq)
296 DBRefEntry[] dbrefs = seq.getDBRefs();
301 for (DBRefEntry dbref : dbrefs)
303 Mapping mapping = dbref.getMap();
304 if (mapping == null || mapping.getTo() == null)
309 SequenceI mapTo = mapping.getTo();
310 MapList map = mapping.getMap();
311 if (map.getFromRatio() == 3)
314 * dna-to-peptide product mapping
316 AlignmentUtils.computeProteinFeatures(seq, mapTo, map);
321 * nucleotide-to-nucleotide mapping e.g. transcript to CDS
323 List<SequenceFeature> features = seq.getFeatures()
324 .getPositionalFeatures(SequenceOntologyI.SEQUENCE_VARIANT);
325 for (SequenceFeature sf : features)
327 if (FEATURE_GROUP_VCF.equals(sf.getFeatureGroup()))
329 transferFeature(sf, mapTo, map);
337 * Tries to add overlapping variants read from a VCF file to the given
338 * sequence, and returns the number of variant features added. Note that this
339 * requires the sequence to hold information as to its chromosomal positions
340 * and reference, in order to be able to map the VCF variants to the sequence.
344 * @param isVcfRefGrch37
347 protected int loadSequenceVCF(SequenceI seq, VCFReader reader,
348 boolean isVcfRefGrch37)
351 GeneLociI seqCoords = seq.getGeneLoci();
352 if (seqCoords == null)
357 List<int[]> seqChromosomalContigs = seqCoords.getMap().getToRanges();
358 for (int[] range : seqChromosomalContigs)
360 count += addVcfVariants(seq, reader, range, isVcfRefGrch37);
367 * Queries the VCF reader for any variants that overlap the given chromosome
368 * region of the sequence, and adds as variant features. Returns the number of
369 * overlapping variants found.
374 * start-end range of a sequence region in its chromosomal
376 * @param isVcfRefGrch37
377 * true if the VCF is with reference to GRCh37
380 protected int addVcfVariants(SequenceI seq, VCFReader reader,
381 int[] range, boolean isVcfRefGrch37)
383 GeneLociI seqCoords = seq.getGeneLoci();
385 String chromosome = seqCoords.getChromosomeId();
386 String seqRef = seqCoords.getAssemblyId();
387 String species = seqCoords.getSpeciesId();
390 * map chromosomal coordinates from GRCh38 (sequence) to
391 * GRCh37 (VCF) if necessary
393 // TODO generalise for other assemblies and species
395 String fromRef = "GRCh38";
396 if (fromRef.equalsIgnoreCase(seqRef) && isVcfRefGrch37)
398 String toRef = "GRCh37";
399 int[] newRange = mapReferenceRange(range, chromosome, "human",
401 if (newRange == null)
403 System.err.println(String.format(
404 "Failed to map %s:%s:%s:%d:%d to %s", species, chromosome,
405 fromRef, range[0], range[1], toRef));
408 offset = newRange[0] - range[0];
412 boolean forwardStrand = range[0] <= range[1];
415 * query the VCF for overlaps
416 * (convert a reverse strand range to forwards)
419 MapList mapping = seqCoords.getMap();
421 int fromLocus = Math.min(range[0], range[1]);
422 int toLocus = Math.max(range[0], range[1]);
423 CloseableIterator<VariantContext> variants = reader.query(chromosome,
425 while (variants.hasNext())
428 * get variant location in sequence chromosomal coordinates
430 VariantContext variant = variants.next();
433 * we can only process SNP variants (which can be reported
434 * as part of a MIXED variant record
436 if (!variant.isSNP() && !variant.isMixed())
441 int start = variant.getStart() - offset;
442 int end = variant.getEnd() - offset;
445 * convert chromosomal location to sequence coordinates
446 * - null if a partially overlapping feature
448 int[] seqLocation = mapping.locateInFrom(start, end);
449 if (seqLocation != null)
451 count += addAlleleFeatures(seq, variant, seqLocation[0],
452 seqLocation[1], forwardStrand);
462 * A convenience method to get the AF value for the given alternate allele
469 protected float getAlleleFrequency(VariantContext variant, int alleleIndex)
472 String attributeValue = getAttributeValue(variant,
473 ALLELE_FREQUENCY_KEY, alleleIndex);
474 if (attributeValue != null)
478 score = Float.parseFloat(attributeValue);
479 } catch (NumberFormatException e)
489 * A convenience method to get an attribute value for an alternate allele
492 * @param attributeName
496 protected String getAttributeValue(VariantContext variant,
497 String attributeName, int alleleIndex)
499 Object att = variant.getAttribute(attributeName);
501 if (att instanceof String)
505 else if (att instanceof ArrayList)
507 return ((List<String>) att).get(alleleIndex);
514 * Adds one variant feature for each allele in the VCF variant record, and
515 * returns the number of features added.
519 * @param featureStart
521 * @param forwardStrand
524 protected int addAlleleFeatures(SequenceI seq, VariantContext variant,
525 int featureStart, int featureEnd, boolean forwardStrand)
530 * Javadoc says getAlternateAlleles() imposes no order on the list returned
531 * so we proceed defensively to get them in strict order
533 int altAlleleCount = variant.getAlternateAlleles().size();
534 for (int i = 0; i < altAlleleCount; i++)
536 added += addAlleleFeature(seq, variant, i, featureStart, featureEnd,
543 * Inspects one allele and attempts to add a variant feature for it to the
544 * sequence. We extract as much as possible of the additional data associated
545 * with this allele to store in the feature's key-value map. Answers the
546 * number of features added (0 or 1).
550 * @param altAlleleIndex
552 * @param featureStart
554 * @param forwardStrand
557 protected int addAlleleFeature(SequenceI seq, VariantContext variant,
558 int altAlleleIndex, int featureStart, int featureEnd,
559 boolean forwardStrand)
561 String reference = variant.getReference().getBaseString();
562 Allele alt = variant.getAlternateAllele(altAlleleIndex);
563 String allele = alt.getBaseString();
564 if (allele.length() != 1)
573 * build the ref,alt allele description e.g. "G,A", using the base
574 * complement if the sequence is on the reverse strand
576 // TODO check how structural variants are shown on reverse strand
577 StringBuilder sb = new StringBuilder();
578 sb.append(forwardStrand ? reference : Dna.reverseComplement(reference));
580 sb.append(forwardStrand ? allele : Dna.reverseComplement(allele));
581 String alleles = sb.toString(); // e.g. G,A
583 String type = SequenceOntologyI.SEQUENCE_VARIANT;
584 float score = getAlleleFrequency(variant, altAlleleIndex);
586 SequenceFeature sf = new SequenceFeature(type, alleles, featureStart,
587 featureEnd, score, FEATURE_GROUP_VCF);
589 sf.setValue(Gff3Helper.ALLELES, alleles);
591 addAlleleProperties(variant, seq, sf, altAlleleIndex);
593 seq.addSequenceFeature(sf);
599 * Add any allele-specific VCF key-value data to the sequence feature
604 * @param altAlelleIndex
607 protected void addAlleleProperties(VariantContext variant, SequenceI seq,
608 SequenceFeature sf, final int altAlelleIndex)
610 Map<String, Object> atts = variant.getAttributes();
612 for (Entry<String, Object> att : atts.entrySet())
614 String key = att.getKey();
617 * extract Consequence data (if present) that we are able to
618 * associated with the allele for this variant feature
622 addConsequences(variant, seq, sf, altAlelleIndex);
627 * we extract values for other data which are allele-specific;
628 * these may be per alternate allele (INFO[key].Number = 'A')
629 * or per allele including reference (INFO[key].Number = 'R')
631 VCFInfoHeaderLine infoHeader = header.getInfoHeaderLine(key);
632 if (infoHeader == null)
635 * can't be sure what data belongs to this allele, so
636 * play safe and don't take any
641 VCFHeaderLineCount number = infoHeader.getCountType();
642 int index = altAlelleIndex;
643 if (number == VCFHeaderLineCount.R)
646 * one value per allele including reference, so bump index
647 * e.g. the 3rd value is for the 2nd alternate allele
651 else if (number != VCFHeaderLineCount.A)
654 * don't save other values as not allele-related
660 * take the index'th value
662 String value = getAttributeValue(variant, key, index);
665 sf.setValue(key, value);
671 * Inspects CSQ data blocks (consequences) and adds attributes on the sequence
672 * feature for the current allele (and transcript if applicable)
674 * Allele matching: if field ALLELE_NUM is present, it must match
675 * altAlleleIndex. If not present, then field Allele value must match the VCF
678 * Transcript matching: if sequence name can be identified to at least one of
679 * the consequences' Feature values, then select only consequences that match
680 * the value (i.e. consequences for the current transcript sequence). If not,
681 * take all consequences (this is the case when adding features to the gene
687 * @param altAlelleIndex
690 protected void addConsequences(VariantContext variant, SequenceI seq,
691 SequenceFeature sf, int altAlelleIndex)
693 Object value = variant.getAttribute(CSQ);
695 if (value == null || !(value instanceof ArrayList<?>))
700 List<String> consequences = (List<String>) value;
703 * if CSQ data includes 'Feature', and any value matches the sequence name,
704 * then restrict consequence data to only the matching value (transcript)
705 * i.e. just pick out consequences for the transcript the variant feature is on
707 String seqName = seq.getName()== null ? "" : seq.getName().toLowerCase();
708 String matchFeature = null;
709 if (csqFeatureFieldIndex > -1)
711 for (String consequence : consequences)
713 String[] csqFields = consequence.split(PIPE_REGEX);
714 if (csqFields.length > csqFeatureFieldIndex)
716 String featureIdentifier = csqFields[csqFeatureFieldIndex];
717 if (featureIdentifier.length() > 4
718 && seqName.indexOf(featureIdentifier.toLowerCase()) > -1)
720 matchFeature = featureIdentifier;
726 StringBuilder sb = new StringBuilder(128);
727 boolean found = false;
729 for (String consequence : consequences)
731 String[] csqFields = consequence.split(PIPE_REGEX);
733 if (includeConsequence(csqFields, matchFeature, variant,
741 sb.append(consequence);
747 sf.setValue(CSQ, sb.toString());
752 * Answers true if we want to associate this block of consequence data with
753 * the specified alternate allele of the VCF variant.
755 * If consequence data includes the ALLELE_NUM field, then this has to match
756 * altAlleleIndex. Otherwise the Allele field of the consequence data has to
757 * match the allele value.
759 * Optionally (if matchFeature is not null), restrict to only include
760 * consequences whose Feature value matches. This allows us to attach
761 * consequences to their respective transcripts.
764 * @param matchFeature
766 * @param altAlelleIndex
770 protected boolean includeConsequence(String[] csqFields,
771 String matchFeature, VariantContext variant, int altAlelleIndex)
774 * check consequence is for the current transcript
776 if (matchFeature != null)
778 if (csqFields.length <= csqFeatureFieldIndex)
782 String featureIdentifier = csqFields[csqFeatureFieldIndex];
783 if (!featureIdentifier.equals(matchFeature))
785 return false; // consequence is for a different transcript
790 * if ALLELE_NUM is present, it must match altAlleleIndex
791 * NB first alternate allele is 1 for ALLELE_NUM, 0 for altAlleleIndex
793 if (csqAlleleNumberFieldIndex > -1)
795 if (csqFields.length <= csqAlleleNumberFieldIndex)
799 String alleleNum = csqFields[csqAlleleNumberFieldIndex];
800 return String.valueOf(altAlelleIndex + 1).equals(alleleNum);
804 * else consequence allele must match variant allele
806 if (csqAlleleFieldIndex > -1 && csqFields.length > csqAlleleFieldIndex)
808 String csqAllele = csqFields[csqAlleleFieldIndex];
809 String vcfAllele = variant.getAlternateAllele(altAlelleIndex)
811 return csqAllele.equals(vcfAllele);
818 * A convenience method to complement a dna base and return the string value
824 protected String complement(byte[] reference)
826 return String.valueOf(Dna.getComplement((char) reference[0]));
830 * Determines the location of the query range (chromosome positions) in a
831 * different reference assembly.
833 * If the range is just a subregion of one for which we already have a mapping
834 * (for example, an exon sub-region of a gene), then the mapping is just
835 * computed arithmetically.
837 * Otherwise, calls the Ensembl REST service that maps from one assembly
838 * reference's coordinates to another's
841 * start-end chromosomal range in 'fromRef' coordinates
845 * assembly reference for the query coordinates
847 * assembly reference we wish to translate to
848 * @return the start-end range in 'toRef' coordinates
850 protected int[] mapReferenceRange(int[] queryRange, String chromosome,
851 String species, String fromRef, String toRef)
854 * first try shorcut of computing the mapping as a subregion of one
855 * we already have (e.g. for an exon, if we have the gene mapping)
857 int[] mappedRange = findSubsumedRangeMapping(queryRange, chromosome,
858 species, fromRef, toRef);
859 if (mappedRange != null)
865 * call (e.g.) http://rest.ensembl.org/map/human/GRCh38/17:45051610..45109016:1/GRCh37
867 EnsemblMap mapper = new EnsemblMap();
868 int[] mapping = mapper.getMapping(species, chromosome, fromRef, toRef,
873 // mapping service failure
878 * save mapping for possible future re-use
880 String key = makeRangesKey(chromosome, species, fromRef, toRef);
881 if (!assemblyMappings.containsKey(key))
883 assemblyMappings.put(key, new HashMap<int[], int[]>());
886 assemblyMappings.get(key).put(queryRange, mapping);
892 * If we already have a 1:1 contiguous mapping which subsumes the given query
893 * range, this method just calculates and returns the subset of that mapping,
894 * else it returns null. In practical terms, if a gene has a contiguous
895 * mapping between (for example) GRCh37 and GRCh38, then we assume that its
896 * subsidiary exons occupy unchanged relative positions, and just compute
897 * these as offsets, rather than do another lookup of the mapping.
899 * If in future these assumptions prove invalid (e.g. for bacterial dna?!),
900 * simply remove this method or let it always return null.
902 * Warning: many rapid calls to the /map service map result in a 429 overload
912 protected int[] findSubsumedRangeMapping(int[] queryRange, String chromosome,
913 String species, String fromRef, String toRef)
915 String key = makeRangesKey(chromosome, species, fromRef, toRef);
916 if (assemblyMappings.containsKey(key))
918 Map<int[], int[]> mappedRanges = assemblyMappings.get(key);
919 for (Entry<int[], int[]> mappedRange : mappedRanges.entrySet())
921 int[] fromRange = mappedRange.getKey();
922 int[] toRange = mappedRange.getValue();
923 if (fromRange[1] - fromRange[0] == toRange[1] - toRange[0])
926 * mapping is 1:1 in length, so we trust it to have no discontinuities
928 if (MappingUtils.rangeContains(fromRange, queryRange))
931 * fromRange subsumes our query range
933 int offset = queryRange[0] - fromRange[0];
934 int mappedRangeFrom = toRange[0] + offset;
935 int mappedRangeTo = mappedRangeFrom + (queryRange[1] - queryRange[0]);
936 return new int[] { mappedRangeFrom, mappedRangeTo };
945 * Transfers the sequence feature to the target sequence, locating its start
946 * and end range based on the mapping. Features which do not overlap the
947 * target sequence are ignored.
950 * @param targetSequence
952 * mapping from the feature's coordinates to the target sequence
954 protected void transferFeature(SequenceFeature sf,
955 SequenceI targetSequence, MapList mapping)
957 int[] mappedRange = mapping.locateInTo(sf.getBegin(), sf.getEnd());
959 if (mappedRange != null)
961 String group = sf.getFeatureGroup();
962 int newBegin = Math.min(mappedRange[0], mappedRange[1]);
963 int newEnd = Math.max(mappedRange[0], mappedRange[1]);
964 SequenceFeature copy = new SequenceFeature(sf, newBegin, newEnd,
965 group, sf.getScore());
966 targetSequence.addSequenceFeature(copy);
971 * Formats a ranges map lookup key
979 protected static String makeRangesKey(String chromosome, String species,
980 String fromRef, String toRef)
982 return species + EXCL + chromosome + EXCL + fromRef + EXCL