badData;
+
/**
* Constructor given a VCF file
*
@@ -245,12 +310,18 @@ public class VCFLoader
*/
public SequenceI loadVCFContig(String contig)
{
- String ref = header.getOtherHeaderLine(VCFHeader.REFERENCE_KEY)
- .getValue();
+ VCFHeaderLine headerLine = header.getOtherHeaderLine(VCFHeader.REFERENCE_KEY);
+ if (headerLine == null)
+ {
+ Cache.log.error("VCF reference header not found");
+ return null;
+ }
+ String ref = headerLine.getValue();
if (ref.startsWith("file://"))
{
ref = ref.substring(7);
}
+ setSpeciesAndAssembly(ref);
SequenceI seq = null;
File dbFile = new File(ref);
@@ -259,12 +330,12 @@ public class VCFLoader
{
HtsContigDb db = new HtsContigDb("", dbFile);
seq = db.getSequenceProxy(contig);
- loadSequenceVCF(seq, ref);
+ loadSequenceVCF(seq);
db.close();
}
else
{
- System.err.println("VCF reference not found: " + ref);
+ Cache.log.error("VCF reference not found: " + ref);
}
return seq;
@@ -283,7 +354,9 @@ public class VCFLoader
{
VCFHeaderLine ref = header
.getOtherHeaderLine(VCFHeader.REFERENCE_KEY);
- String vcfAssembly = ref.getValue();
+ String reference = ref == null ? null : ref.getValue();
+
+ setSpeciesAndAssembly(reference);
int varCount = 0;
int seqCount = 0;
@@ -293,7 +366,7 @@ public class VCFLoader
*/
for (SequenceI seq : seqs)
{
- int added = loadSequenceVCF(seq, vcfAssembly);
+ int added = loadSequenceVCF(seq);
if (added > 0)
{
seqCount++;
@@ -337,6 +410,71 @@ public class VCFLoader
}
/**
+ * Attempts to determine and save the species and genome assembly version to
+ * which the VCF data applies. This may be done by parsing the {@code reference}
+ * header line, configured in a property file, or (potentially) confirmed
+ * interactively by the user.
+ *
+ * The saved values should be identifiers valid for Ensembl's REST service
+ * {@code map} endpoint, so they can be used (if necessary) to retrieve the
+ * mapping between VCF coordinates and sequence coordinates.
+ *
+ * @param reference
+ * @see https://rest.ensembl.org/documentation/info/assembly_map
+ * @see https://rest.ensembl.org/info/assembly/human?content-type=text/xml
+ * @see https://rest.ensembl.org/info/species?content-type=text/xml
+ */
+ protected void setSpeciesAndAssembly(String reference)
+ {
+ if (reference == null)
+ {
+ Cache.log.error("No VCF ##reference found, defaulting to "
+ + DEFAULT_REFERENCE + ":" + DEFAULT_SPECIES);
+ reference = DEFAULT_REFERENCE; // default to GRCh37 if not specified
+ }
+ reference = reference.toLowerCase();
+
+ /*
+ * for a non-human species, or other assembly identifier,
+ * specify as a Jalview property file entry e.g.
+ * VCF_ASSEMBLY = hs37=GRCh37,assembly19=GRCh37
+ * VCF_SPECIES = c_elegans=celegans
+ * to map a token in the reference header to a value
+ */
+ String prop = Cache.getDefault(VCF_ASSEMBLY, DEFAULT_VCF_ASSEMBLY);
+ for (String token : prop.split(","))
+ {
+ String[] tokens = token.split("=");
+ if (tokens.length == 2)
+ {
+ if (reference.contains(tokens[0].trim().toLowerCase()))
+ {
+ vcfAssembly = tokens[1].trim();
+ break;
+ }
+ }
+ }
+
+ vcfSpecies = DEFAULT_SPECIES;
+ prop = Cache.getProperty(VCF_SPECIES);
+ if (prop != null)
+ {
+ for (String token : prop.split(","))
+ {
+ String[] tokens = token.split("=");
+ if (tokens.length == 2)
+ {
+ if (reference.contains(tokens[0].trim().toLowerCase()))
+ {
+ vcfSpecies = tokens[1].trim();
+ break;
+ }
+ }
+ }
+ }
+ }
+
+ /**
* Opens the VCF file and parses header data
*
* @param filePath
@@ -588,12 +726,11 @@ public class VCFLoader
* and returns the number of variant features added
*
* @param seq
- * @param vcfAssembly
* @return
*/
- protected int loadSequenceVCF(SequenceI seq, String vcfAssembly)
+ protected int loadSequenceVCF(SequenceI seq)
{
- VCFMap vcfMap = getVcfMap(seq, vcfAssembly);
+ VCFMap vcfMap = getVcfMap(seq);
if (vcfMap == null)
{
return 0;
@@ -614,10 +751,9 @@ public class VCFLoader
* Answers a map from sequence coordinates to VCF chromosome ranges
*
* @param seq
- * @param vcfAssembly
* @return
*/
- private VCFMap getVcfMap(SequenceI seq, String vcfAssembly)
+ private VCFMap getVcfMap(SequenceI seq)
{
/*
* simplest case: sequence has id and length matching a VCF contig
@@ -648,34 +784,28 @@ public class VCFLoader
String species = seqCoords.getSpeciesId();
String chromosome = seqCoords.getChromosomeId();
String seqRef = seqCoords.getAssemblyId();
- MapList map = seqCoords.getMap();
+ MapList map = seqCoords.getMapping();
- if (!vcfSpeciesMatchesSequence(vcfAssembly, species))
+ // note this requires the configured species to match that
+ // returned with the Ensembl sequence; todo: support aliases?
+ if (!vcfSpecies.equalsIgnoreCase(species))
{
+ Cache.log.warn("No VCF loaded to " + seq.getName()
+ + " as species not matched");
return null;
}
- if (vcfAssemblyMatchesSequence(vcfAssembly, seqRef))
+ if (seqRef.equalsIgnoreCase(vcfAssembly))
{
return new VCFMap(chromosome, map);
}
- if (!"GRCh38".equalsIgnoreCase(seqRef) // Ensembl
- || !vcfAssembly.contains("Homo_sapiens_assembly19")) // gnomAD
- {
- return null;
- }
-
/*
- * map chromosomal coordinates from sequence to VCF if the VCF
- * data has a different reference assembly to the sequence
+ * VCF data has a different reference assembly to the sequence:
+ * query Ensembl to map chromosomal coordinates from sequence to VCF
*/
- // TODO generalise for cases other than GRCh38 -> GRCh37 !
- // - or get the user to choose in a dialog
-
List toVcfRanges = new ArrayList<>();
List fromSequenceRanges = new ArrayList<>();
- String toRef = "GRCh37";
for (int[] range : map.getToRanges())
{
@@ -687,12 +817,13 @@ public class VCFLoader
}
int[] newRange = mapReferenceRange(range, chromosome, "human", seqRef,
- toRef);
+ vcfAssembly);
if (newRange == null)
{
Cache.log.error(
String.format("Failed to map %s:%s:%s:%d:%d to %s", species,
- chromosome, seqRef, range[0], range[1], toRef));
+ chromosome, seqRef, range[0], range[1],
+ vcfAssembly));
continue;
}
else
@@ -733,62 +864,6 @@ public class VCFLoader
}
/**
- * Answers true if we determine that the VCF data uses the same reference
- * assembly as the sequence, else false
- *
- * @param vcfAssembly
- * @param seqRef
- * @return
- */
- private boolean vcfAssemblyMatchesSequence(String vcfAssembly,
- String seqRef)
- {
- // TODO improve on this stub, which handles gnomAD and
- // hopes for the best for other cases
-
- if ("GRCh38".equalsIgnoreCase(seqRef) // Ensembl
- && vcfAssembly.contains("Homo_sapiens_assembly19")) // gnomAD
- {
- return false;
- }
- return true;
- }
-
- /**
- * Answers true if the species inferred from the VCF reference identifier
- * matches that for the sequence
- *
- * @param vcfAssembly
- * @param speciesId
- * @return
- */
- boolean vcfSpeciesMatchesSequence(String vcfAssembly, String speciesId)
- {
- // PROBLEM 1
- // there are many aliases for species - how to equate one with another?
- // PROBLEM 2
- // VCF ##reference header is an unstructured URI - how to extract species?
- // perhaps check if ref includes any (Ensembl) alias of speciesId??
- // TODO ask the user to confirm this??
-
- if (vcfAssembly.contains("Homo_sapiens") // gnomAD exome data example
- && "HOMO_SAPIENS".equals(speciesId)) // Ensembl species id
- {
- return true;
- }
-
- if (vcfAssembly.contains("c_elegans") // VEP VCF response example
- && "CAENORHABDITIS_ELEGANS".equals(speciesId)) // Ensembl
- {
- return true;
- }
-
- // this is not a sustainable solution...
-
- return false;
- }
-
- /**
* Queries the VCF reader for any variants that overlap the mapped chromosome
* ranges of the sequence, and adds as variant features. Returns the number of
* overlapping variants found.
@@ -811,54 +886,38 @@ public class VCFLoader
{
int vcfStart = Math.min(range[0], range[1]);
int vcfEnd = Math.max(range[0], range[1]);
- CloseableIterator variants = reader
- .query(map.chromosome, vcfStart, vcfEnd);
- while (variants.hasNext())
+ try
{
- VariantContext variant = variants.next();
-
- int[] featureRange = map.map.locateInFrom(variant.getStart(),
- variant.getEnd());
-
- if (featureRange != null)
+ CloseableIterator variants = reader
+ .query(map.chromosome, vcfStart, vcfEnd);
+ while (variants.hasNext())
{
- int featureStart = Math.min(featureRange[0], featureRange[1]);
- int featureEnd = Math.max(featureRange[0], featureRange[1]);
- count += addAlleleFeatures(seq, variant, featureStart, featureEnd,
- forwardStrand);
- }
- }
- variants.close();
- }
+ VariantContext variant = variants.next();
- return count;
- }
+ int[] featureRange = map.map.locateInFrom(variant.getStart(),
+ variant.getEnd());
- /**
- * A convenience method to get the AF value for the given alternate allele
- * index
- *
- * @param variant
- * @param alleleIndex
- * @return
- */
- protected float getAlleleFrequency(VariantContext variant, int alleleIndex)
- {
- float score = 0f;
- String attributeValue = getAttributeValue(variant,
- ALLELE_FREQUENCY_KEY, alleleIndex);
- if (attributeValue != null)
- {
- try
- {
- score = Float.parseFloat(attributeValue);
- } catch (NumberFormatException e)
+ if (featureRange != null)
+ {
+ int featureStart = Math.min(featureRange[0], featureRange[1]);
+ int featureEnd = Math.max(featureRange[0], featureRange[1]);
+ count += addAlleleFeatures(seq, variant, featureStart,
+ featureEnd, forwardStrand);
+ }
+ }
+ variants.close();
+ } catch (TribbleException e)
{
- // leave as 0
+ /*
+ * RuntimeException throwable by htsjdk
+ */
+ String msg = String.format("Error reading VCF for %s:%d-%d: %s ",
+ map.chromosome, vcfStart, vcfEnd);
+ Cache.log.error(msg);
}
}
- return score;
+ return count;
}
/**
@@ -981,13 +1040,24 @@ public class VCFLoader
type = getOntologyTerm(consequence);
}
- float score = getAlleleFrequency(variant, altAlleleIndex);
-
SequenceFeature sf = new SequenceFeature(type, alleles, featureStart,
- featureEnd, score, FEATURE_GROUP_VCF);
+ featureEnd, FEATURE_GROUP_VCF);
sf.setSource(sourceId);
- sf.setValue(Gff3Helper.ALLELES, alleles);
+ /*
+ * save the derived alleles as a named attribute; this will be
+ * needed when Jalview computes derived peptide variants
+ */
+ addFeatureAttribute(sf, Gff3Helper.ALLELES, alleles);
+
+ /*
+ * add selected VCF fixed column data as feature attributes
+ */
+ addFeatureAttribute(sf, VCF_POS, String.valueOf(variant.getStart()));
+ addFeatureAttribute(sf, VCF_ID, variant.getID());
+ addFeatureAttribute(sf, VCF_QUAL,
+ String.valueOf(variant.getPhredScaledQual()));
+ addFeatureAttribute(sf, VCF_FILTER, getFilter(variant));
addAlleleProperties(variant, sf, altAlleleIndex, consequence);
@@ -997,6 +1067,53 @@ public class VCFLoader
}
/**
+ * Answers the VCF FILTER value for the variant - or an approximation to it.
+ * This field is either PASS, or a semi-colon separated list of filters not
+ * passed. htsjdk saves filters as a HashSet, so the order when reassembled into
+ * a list may be different.
+ *
+ * @param variant
+ * @return
+ */
+ String getFilter(VariantContext variant)
+ {
+ Set filters = variant.getFilters();
+ if (filters.isEmpty())
+ {
+ return NO_VALUE;
+ }
+ Iterator iterator = filters.iterator();
+ String first = iterator.next();
+ if (filters.size() == 1)
+ {
+ return first;
+ }
+
+ StringBuilder sb = new StringBuilder(first);
+ while (iterator.hasNext())
+ {
+ sb.append(";").append(iterator.next());
+ }
+
+ return sb.toString();
+ }
+
+ /**
+ * Adds one feature attribute unless the value is null, empty or '.'
+ *
+ * @param sf
+ * @param key
+ * @param value
+ */
+ void addFeatureAttribute(SequenceFeature sf, String key, String value)
+ {
+ if (value != null && !value.isEmpty() && !NO_VALUE.equals(value))
+ {
+ sf.setValue(key, value);
+ }
+ }
+
+ /**
* Determines the Sequence Ontology term to use for the variant feature type in
* Jalview. The default is 'sequence_variant', but a more specific term is used
* if:
@@ -1190,14 +1307,6 @@ public class VCFLoader
}
/*
- * filter out fields we don't want to capture
- */
- if (!vcfFieldsOfInterest.contains(key))
- {
- continue;
- }
-
- /*
* we extract values for other data which are allele-specific;
* these may be per alternate allele (INFO[key].Number = 'A')
* or per allele including reference (INFO[key].Number = 'R')
@@ -1234,10 +1343,81 @@ public class VCFLoader
* take the index'th value
*/
String value = getAttributeValue(variant, key, index);
- if (value != null)
+ if (value != null && isValid(variant, key, value))
+ {
+ /*
+ * decode colon, semicolon, equals sign, percent sign, comma (only)
+ * as required by the VCF specification (para 1.2)
+ */
+ value = StringUtils.urlDecode(value, VCF_ENCODABLE);
+ addFeatureAttribute(sf, key, value);
+ }
+ }
+ }
+
+ /**
+ * Answers true for '.', null, or an empty value, or if the INFO type is String.
+ * If the INFO type is Integer or Float, answers false if the value is not in
+ * valid format.
+ *
+ * @param variant
+ * @param infoId
+ * @param value
+ * @return
+ */
+ protected boolean isValid(VariantContext variant, String infoId,
+ String value)
+ {
+ if (value == null || value.isEmpty() || NO_VALUE.equals(value))
+ {
+ return true;
+ }
+ VCFInfoHeaderLine infoHeader = header.getInfoHeaderLine(infoId);
+ if (infoHeader == null)
+ {
+ Cache.log.error("Field " + infoId + " has no INFO header");
+ return false;
+ }
+ VCFHeaderLineType infoType = infoHeader.getType();
+ try
+ {
+ if (infoType == VCFHeaderLineType.Integer)
+ {
+ Integer.parseInt(value);
+ }
+ else if (infoType == VCFHeaderLineType.Float)
{
- sf.setValue(key, value);
+ Float.parseFloat(value);
}
+ } catch (NumberFormatException e)
+ {
+ logInvalidValue(variant, infoId, value);
+ return false;
+ }
+ return true;
+ }
+
+ /**
+ * Logs an error message for malformed data; duplicate messages (same id and
+ * value) are not logged
+ *
+ * @param variant
+ * @param infoId
+ * @param value
+ */
+ private void logInvalidValue(VariantContext variant, String infoId,
+ String value)
+ {
+ if (badData == null)
+ {
+ badData = new HashSet<>();
+ }
+ String token = infoId + ":" + value;
+ if (!badData.contains(token))
+ {
+ badData.add(token);
+ Cache.log.error(String.format("Invalid VCF data at %s:%d %s=%s",
+ variant.getContig(), variant.getStart(), infoId, value));
}
}
@@ -1289,6 +1469,11 @@ public class VCFLoader
String id = vepFieldsOfInterest.get(i);
if (id != null)
{
+ /*
+ * VCF spec requires encoding of special characters e.g. '='
+ * so decode them here before storing
+ */
+ field = StringUtils.urlDecode(field, VCF_ENCODABLE);
csqValues.put(id, field);
}
}