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_NUM_KEY = "ALLELE_NUM"; // 0 (ref), 1...
49 private static final String FEATURE_KEY = "Feature"; // Ensembl stable id
52 * default VCF INFO key for VEP consequence data
53 * NB this can be overridden running VEP with --vcf_info_field
54 * - we don't handle this case (require CSQ identifier)
56 private static final String CSQ = "CSQ";
59 * separator for fields in consequence data
61 private static final String PIPE = "|";
63 private static final String PIPE_REGEX = "\\" + PIPE;
66 * key for Allele Frequency output by VEP
67 * see http://www.ensembl.org/info/docs/tools/vep/vep_formats.html
69 private static final String ALLELE_FREQUENCY_KEY = "AF";
72 * delimiter that separates multiple consequence data blocks
74 private static final String COMMA = ",";
77 * (temporary) flag that determines whether Jalview adds one feature
78 * per VCF record, or one per allele (preferred)
80 private static final boolean FEATURE_PER_ALLELE = true;
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 the ALLELE_NUM 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 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 * note offset of CSQ ALLELE_NUM field if it is declared
184 findAlleleNumberFieldIndex();
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 = loadVCF(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 * If the CSQ INFO header declares that ALLELE_NUM is included in the data,
243 * record its (pipe-delimited) offset in each (comma-delimited) consequence
244 * block; CSQ fields are declared in the CSQ INFO Description e.g.
246 * Description="Consequence ...from ... VEP. Format: Allele|Consequence|...
248 protected void findAlleleNumberFieldIndex()
250 VCFInfoHeaderLine csqInfo = header.getInfoHeaderLine(CSQ);
251 String desc = csqInfo.getDescription();
254 String[] format = desc.split(PIPE_REGEX);
256 for (String field : format)
258 if (ALLELE_NUM_KEY.equals(field))
260 csqAlleleNumberFieldIndex = index;
262 if (FEATURE_KEY.equals(field))
264 csqFeatureFieldIndex = index;
272 * Transfers VCF features to sequences to which this sequence has a mapping.
273 * If the mapping is 1:3, computes peptide variants from nucleotide variants.
277 protected void transferAddedFeatures(SequenceI seq)
279 DBRefEntry[] dbrefs = seq.getDBRefs();
284 for (DBRefEntry dbref : dbrefs)
286 Mapping mapping = dbref.getMap();
287 if (mapping == null || mapping.getTo() == null)
292 SequenceI mapTo = mapping.getTo();
293 MapList map = mapping.getMap();
294 if (map.getFromRatio() == 3)
297 * dna-to-peptide product mapping
299 AlignmentUtils.computeProteinFeatures(seq, mapTo, map);
304 * nucleotide-to-nucleotide mapping e.g. transcript to CDS
306 // TODO no DBRef to CDS is added to transcripts
307 List<SequenceFeature> features = seq.getFeatures()
308 .getPositionalFeatures(SequenceOntologyI.SEQUENCE_VARIANT);
309 for (SequenceFeature sf : features)
311 if (FEATURE_GROUP_VCF.equals(sf.getFeatureGroup()))
313 transferFeature(sf, mapTo, map);
321 * Tries to add overlapping variants read from a VCF file to the given
322 * sequence, and returns the number of variant features added. Note that this
323 * requires the sequence to hold information as to its chromosomal positions
324 * and reference, in order to be able to map the VCF variants to the sequence.
328 * @param isVcfRefGrch37
331 protected int loadVCF(SequenceI seq, VCFReader reader,
332 boolean isVcfRefGrch37)
335 GeneLociI seqCoords = seq.getGeneLoci();
336 if (seqCoords == null)
341 List<int[]> seqChromosomalContigs = seqCoords.getMap().getToRanges();
342 for (int[] range : seqChromosomalContigs)
344 count += addVcfVariants(seq, reader, range, isVcfRefGrch37);
351 * Queries the VCF reader for any variants that overlap the given chromosome
352 * region of the sequence, and adds as variant features. Returns the number of
353 * overlapping variants found.
358 * start-end range of a sequence region in its chromosomal
360 * @param isVcfRefGrch37
361 * true if the VCF is with reference to GRCh37
364 protected int addVcfVariants(SequenceI seq, VCFReader reader,
365 int[] range, boolean isVcfRefGrch37)
367 GeneLociI seqCoords = seq.getGeneLoci();
369 String chromosome = seqCoords.getChromosomeId();
370 String seqRef = seqCoords.getAssemblyId();
371 String species = seqCoords.getSpeciesId();
374 * map chromosomal coordinates from GRCh38 (sequence) to
375 * GRCh37 (VCF) if necessary
377 // TODO generalise for other assemblies and species
379 String fromRef = "GRCh38";
380 if (fromRef.equalsIgnoreCase(seqRef) && isVcfRefGrch37)
382 String toRef = "GRCh37";
383 int[] newRange = mapReferenceRange(range, chromosome, "human",
385 if (newRange == null)
387 System.err.println(String.format(
388 "Failed to map %s:%s:%s:%d:%d to %s", species, chromosome,
389 fromRef, range[0], range[1], toRef));
392 offset = newRange[0] - range[0];
396 boolean forwardStrand = range[0] <= range[1];
399 * query the VCF for overlaps
400 * (convert a reverse strand range to forwards)
403 MapList mapping = seqCoords.getMap();
405 int fromLocus = Math.min(range[0], range[1]);
406 int toLocus = Math.max(range[0], range[1]);
407 CloseableIterator<VariantContext> variants = reader.query(chromosome,
409 while (variants.hasNext())
412 * get variant location in sequence chromosomal coordinates
414 VariantContext variant = variants.next();
417 * we can only process SNP variants (which can be reported
418 * as part of a MIXED variant record
420 if (!variant.isSNP() && !variant.isMixed())
425 int start = variant.getStart() - offset;
426 int end = variant.getEnd() - offset;
429 * convert chromosomal location to sequence coordinates
430 * - null if a partially overlapping feature
432 int[] seqLocation = mapping.locateInFrom(start, end);
433 if (seqLocation != null)
435 count += addVariantFeature(seq, variant, seqLocation[0],
436 seqLocation[1], forwardStrand);
446 * Inspects the VCF variant record, and adds variant features to the sequence.
447 * Only SNP variants are added, not INDELs. Returns the number of features
450 * If the sequence maps to the reverse strand of the chromosome, reference and
451 * variant bases are recorded as their complements (C/G, A/T).
455 * @param featureStart
457 * @param forwardStrand
459 protected int addVariantFeature(SequenceI seq, VariantContext variant,
460 int featureStart, int featureEnd, boolean forwardStrand)
462 byte[] reference = variant.getReference().getBases();
463 if (reference.length != 1)
466 * sorry, we don't handle INDEL variants
471 if (FEATURE_PER_ALLELE)
473 return addAlleleFeatures(seq, variant, featureStart, featureEnd,
478 * for now we extract allele frequency as feature score; note
479 * this attribute is String for a simple SNP, but List<String> if
480 * multiple alleles at the locus; we extract for the simple case only
482 float score = getAlleleFrequency(variant, 0);
484 StringBuilder sb = new StringBuilder();
485 sb.append(forwardStrand ? (char) reference[0] : complement(reference));
488 * inspect alleles and record SNP variants (as the variant
489 * record could be MIXED and include INDEL and SNP alleles)
490 * warning: getAlleles gives no guarantee as to the order
491 * in which they are returned
493 for (Allele allele : variant.getAlleles())
495 if (!allele.isReference())
497 byte[] alleleBase = allele.getBases();
498 if (alleleBase.length == 1)
500 sb.append(COMMA).append(
501 forwardStrand ? (char) alleleBase[0]
502 : complement(alleleBase));
506 String alleles = sb.toString(); // e.g. G,A,C
508 String type = SequenceOntologyI.SEQUENCE_VARIANT;
510 SequenceFeature sf = new SequenceFeature(type, alleles, featureStart,
511 featureEnd, score, FEATURE_GROUP_VCF);
513 sf.setValue(Gff3Helper.ALLELES, alleles);
515 Map<String, Object> atts = variant.getAttributes();
516 for (Entry<String, Object> att : atts.entrySet())
518 sf.setValue(att.getKey(), att.getValue());
520 seq.addSequenceFeature(sf);
526 * A convenience method to get the AF value for the given alternate allele
533 protected float getAlleleFrequency(VariantContext variant, int alleleIndex)
536 String attributeValue = getAttributeValue(variant,
537 ALLELE_FREQUENCY_KEY, alleleIndex);
538 if (attributeValue != null)
542 score = Float.parseFloat(attributeValue);
543 } catch (NumberFormatException e)
553 * A convenience method to get an attribute value for an alternate allele
556 * @param attributeName
560 protected String getAttributeValue(VariantContext variant,
561 String attributeName, int alleleIndex)
563 Object att = variant.getAttribute(attributeName);
565 if (att instanceof String)
569 else if (att instanceof ArrayList)
571 return ((List<String>) att).get(alleleIndex);
578 * Adds one variant feature for each SNP allele in the VCF variant record, and
579 * returns the number of features added.
583 * @param featureStart
585 * @param forwardStrand
588 protected int addAlleleFeatures(SequenceI seq, VariantContext variant,
589 int featureStart, int featureEnd, boolean forwardStrand)
594 * Javadoc says getAlternateAlleles() imposes no order on the list returned
595 * so we proceed defensively to get them in strict order
597 int altAlleleCount = variant.getAlternateAlleles().size();
598 for (int i = 0; i < altAlleleCount; i++)
600 added += addAlleleFeature(seq, variant, i, featureStart, featureEnd,
607 * Inspects one allele and attempts to add a variant feature for it to the
608 * sequence. Only SNP variants are added as features. We extract as much as
609 * possible of the additional data associated with this allele to store in the
610 * feature's key-value map. Answers the number of features added (0 or 1).
614 * @param altAlleleIndex
615 * @param featureStart
617 * @param forwardStrand
620 protected int addAlleleFeature(SequenceI seq, VariantContext variant,
621 int altAlleleIndex, int featureStart, int featureEnd,
622 boolean forwardStrand)
624 byte[] reference = variant.getReference().getBases();
625 Allele alt = variant.getAlternateAllele(altAlleleIndex);
626 byte[] allele = alt.getBases();
627 if (allele.length != 1)
636 * build the ref,alt allele description e.g. "G,A"
638 StringBuilder sb = new StringBuilder();
639 sb.append(forwardStrand ? (char) reference[0] : complement(reference));
641 sb.append(forwardStrand ? (char) allele[0] : complement(allele));
642 String alleles = sb.toString(); // e.g. G,A
644 String type = SequenceOntologyI.SEQUENCE_VARIANT;
645 float score = getAlleleFrequency(variant, altAlleleIndex);
647 SequenceFeature sf = new SequenceFeature(type, alleles, featureStart,
648 featureEnd, score, FEATURE_GROUP_VCF);
650 sf.setValue(Gff3Helper.ALLELES, alleles);
652 addAlleleProperties(variant, seq, sf, altAlleleIndex);
654 seq.addSequenceFeature(sf);
660 * Add any allele-specific VCF key-value data to the sequence feature
665 * @param altAlelleIndex
667 protected void addAlleleProperties(VariantContext variant,
669 SequenceFeature sf, final int altAlelleIndex)
671 Map<String, Object> atts = variant.getAttributes();
673 for (Entry<String, Object> att : atts.entrySet())
675 String key = att.getKey();
678 * extract Consequence data (if present) that we are able to
679 * associated with the allele for this variant feature
681 if (CSQ.equals(key) && csqAlleleNumberFieldIndex > -1)
683 addConsequences(att.getValue(), seq, sf, altAlelleIndex + 1);
688 * we extract values for other data which are allele-specific;
689 * these may be per alternate allele (INFO[key].Number = 'A')
690 * or per allele including reference (INFO[key].Number = 'R')
692 VCFHeaderLineCount number = header.getInfoHeaderLine(key)
694 int index = altAlelleIndex;
695 if (number == VCFHeaderLineCount.R)
698 * one value per allele including reference, so bump index
699 * e.g. the 3rd value is for the 2nd alternate allele
703 else if (number != VCFHeaderLineCount.A)
706 * don't save other values as not allele-related
712 * take the index'th value
714 String value = getAttributeValue(variant, key, index);
717 sf.setValue(key, value);
723 * Inspects CSQ data blocks (consequences) and adds attributes on the sequence
724 * feature for the current allele (and transcript if applicable)
726 * Allele matching: we require field ALLELE_NUM to match altAlleleIndex. If
727 * the CSQ data does not include ALLELE_NUM values then no data is added to
728 * the variant feature.
730 * Transcript matching: if sequence name can be identified to at least one of
731 * the consequences' Feature values, then select only consequences that match
732 * the value (i.e. consequences for the current transcript sequence). If not,
733 * take all consequences (this is the case when adding features to the gene
739 * @param altAlelleIndex
740 * (1=first alternative allele...)
742 protected void addConsequences(Object value, SequenceI seq,
746 if (!(value instanceof ArrayList<?>))
751 List<String> consequences = (List<String>) value;
754 * if CSQ data includes 'Feature', and any value matches the sequence name,
755 * then restrict consequence data to the matching value (transcript)
756 * i.e. just pick out consequences for the transcript the variant feature is on
758 String seqName = seq.getName()== null ? "" : seq.getName().toLowerCase();
759 boolean matchFeature = false;
760 String matchFeatureValue = null;
761 if (csqFeatureFieldIndex > -1)
763 for (String consequence : consequences)
765 String[] csqFields = consequence.split(PIPE_REGEX);
766 if (csqFields.length > csqFeatureFieldIndex)
768 String featureIdentifier = csqFields[csqFeatureFieldIndex];
769 if (featureIdentifier.length() > 4
770 && seqName.indexOf(featureIdentifier.toLowerCase()) > -1)
773 matchFeatureValue = featureIdentifier;
779 StringBuilder sb = new StringBuilder(128);
780 boolean found = false;
782 for (String consequence : consequences)
784 String[] csqFields = consequence.split(PIPE_REGEX);
787 * check consequence is for the current transcript
791 if (csqFields.length <= csqFeatureFieldIndex)
795 String featureIdentifier = csqFields[csqFeatureFieldIndex];
796 if (!featureIdentifier.equals(matchFeatureValue))
798 continue; // consequence is for a different transcript
802 if (csqFields.length > csqAlleleNumberFieldIndex)
804 String alleleNum = csqFields[csqAlleleNumberFieldIndex];
805 if (String.valueOf(altAlelleIndex).equals(alleleNum))
812 sb.append(consequence);
819 sf.setValue(CSQ, sb.toString());
824 * A convenience method to complement a dna base and return the string value
830 protected String complement(byte[] reference)
832 return String.valueOf(Dna.getComplement((char) reference[0]));
836 * Determines the location of the query range (chromosome positions) in a
837 * different reference assembly.
839 * If the range is just a subregion of one for which we already have a mapping
840 * (for example, an exon sub-region of a gene), then the mapping is just
841 * computed arithmetically.
843 * Otherwise, calls the Ensembl REST service that maps from one assembly
844 * reference's coordinates to another's
847 * start-end chromosomal range in 'fromRef' coordinates
851 * assembly reference for the query coordinates
853 * assembly reference we wish to translate to
854 * @return the start-end range in 'toRef' coordinates
856 protected int[] mapReferenceRange(int[] queryRange, String chromosome,
857 String species, String fromRef, String toRef)
860 * first try shorcut of computing the mapping as a subregion of one
861 * we already have (e.g. for an exon, if we have the gene mapping)
863 int[] mappedRange = findSubsumedRangeMapping(queryRange, chromosome,
864 species, fromRef, toRef);
865 if (mappedRange != null)
871 * call (e.g.) http://rest.ensembl.org/map/human/GRCh38/17:45051610..45109016:1/GRCh37
873 EnsemblMap mapper = new EnsemblMap();
874 int[] mapping = mapper.getMapping(species, chromosome, fromRef, toRef,
879 // mapping service failure
884 * save mapping for possible future re-use
886 String key = makeRangesKey(chromosome, species, fromRef, toRef);
887 if (!assemblyMappings.containsKey(key))
889 assemblyMappings.put(key, new HashMap<int[], int[]>());
892 assemblyMappings.get(key).put(queryRange, mapping);
898 * If we already have a 1:1 contiguous mapping which subsumes the given query
899 * range, this method just calculates and returns the subset of that mapping,
900 * else it returns null. In practical terms, if a gene has a contiguous
901 * mapping between (for example) GRCh37 and GRCh38, then we assume that its
902 * subsidiary exons occupy unchanged relative positions, and just compute
903 * these as offsets, rather than do another lookup of the mapping.
905 * If in future these assumptions prove invalid (e.g. for bacterial dna?!),
906 * simply remove this method or let it always return null.
908 * Warning: many rapid calls to the /map service map result in a 429 overload
918 protected int[] findSubsumedRangeMapping(int[] queryRange, String chromosome,
919 String species, String fromRef, String toRef)
921 String key = makeRangesKey(chromosome, species, fromRef, toRef);
922 if (assemblyMappings.containsKey(key))
924 Map<int[], int[]> mappedRanges = assemblyMappings.get(key);
925 for (Entry<int[], int[]> mappedRange : mappedRanges.entrySet())
927 int[] fromRange = mappedRange.getKey();
928 int[] toRange = mappedRange.getValue();
929 if (fromRange[1] - fromRange[0] == toRange[1] - toRange[0])
932 * mapping is 1:1 in length, so we trust it to have no discontinuities
934 if (MappingUtils.rangeContains(fromRange, queryRange))
937 * fromRange subsumes our query range
939 int offset = queryRange[0] - fromRange[0];
940 int mappedRangeFrom = toRange[0] + offset;
941 int mappedRangeTo = mappedRangeFrom + (queryRange[1] - queryRange[0]);
942 return new int[] { mappedRangeFrom, mappedRangeTo };
951 * Transfers the sequence feature to the target sequence, locating its start
952 * and end range based on the mapping. Features which do not overlap the
953 * target sequence are ignored.
956 * @param targetSequence
958 * mapping from the feature's coordinates to the target sequence
960 protected void transferFeature(SequenceFeature sf,
961 SequenceI targetSequence, MapList mapping)
963 int[] mappedRange = mapping.locateInTo(sf.getBegin(), sf.getEnd());
965 if (mappedRange != null)
967 String group = sf.getFeatureGroup();
968 int newBegin = Math.min(mappedRange[0], mappedRange[1]);
969 int newEnd = Math.max(mappedRange[0], mappedRange[1]);
970 SequenceFeature copy = new SequenceFeature(sf, newBegin, newEnd,
971 group, sf.getScore());
972 targetSequence.addSequenceFeature(copy);
977 * Formats a ranges map lookup key
985 protected static String makeRangesKey(String chromosome, String species,
986 String fromRef, String toRef)
988 return species + EXCL + chromosome + EXCL + fromRef + EXCL