@Test(dataProvider = "mergeFiltered")
public void testMergeFiltered(MergeFilteredTest cfg) {
final List<String> priority = vcs2priority(cfg.inputs);
final VariantContext merged =
VariantContextUtils.simpleMerge(
genomeLocParser,
cfg.inputs,
priority,
cfg.type,
VariantContextUtils.GenotypeMergeType.PRIORITIZE,
true,
false,
"set",
false,
false);
// test alleles are equal
Assert.assertEquals(merged.getAlleles(), cfg.expected.getAlleles());
// test set field
Assert.assertEquals(merged.getAttribute("set"), cfg.setExpected);
// test filter field
Assert.assertEquals(merged.getFilters(), cfg.expected.getFilters());
}
/**
* add a record to the file
*
* @param vc the Variant Context object
* @param refBase the ref base used for indels
* @param refBaseShouldBeAppliedToEndOfAlleles *** THIS SHOULD BE FALSE EXCEPT FOR AN INDEL AT THE
* EXTREME BEGINNING OF A CONTIG (WHERE THERE IS NO PREVIOUS BASE, SO WE USE THE BASE AFTER
* THE EVENT INSTEAD)
*/
public void add(VariantContext vc, byte refBase, boolean refBaseShouldBeAppliedToEndOfAlleles) {
if (mHeader == null)
throw new IllegalStateException(
"The VCF Header must be written before records can be added: " + locationString());
if (doNotWriteGenotypes) vc = VariantContext.modifyGenotypes(vc, null);
try {
vc =
VariantContext.createVariantContextWithPaddedAlleles(
vc, refBase, refBaseShouldBeAppliedToEndOfAlleles);
// if we are doing on the fly indexing, add the record ***before*** we write any bytes
if (indexer != null) indexer.addFeature(vc, positionalStream.getPosition());
Map<Allele, String> alleleMap = new HashMap<Allele, String>(vc.getAlleles().size());
alleleMap.put(Allele.NO_CALL, VCFConstants.EMPTY_ALLELE); // convenience for lookup
// CHROM
mWriter.write(vc.getChr());
mWriter.write(VCFConstants.FIELD_SEPARATOR);
// POS
mWriter.write(String.valueOf(vc.getStart()));
mWriter.write(VCFConstants.FIELD_SEPARATOR);
// ID
String ID = vc.hasID() ? vc.getID() : VCFConstants.EMPTY_ID_FIELD;
mWriter.write(ID);
mWriter.write(VCFConstants.FIELD_SEPARATOR);
// REF
alleleMap.put(vc.getReference(), "0");
String refString = vc.getReference().getDisplayString();
mWriter.write(refString);
mWriter.write(VCFConstants.FIELD_SEPARATOR);
// ALT
if (vc.isVariant()) {
Allele altAllele = vc.getAlternateAllele(0);
alleleMap.put(altAllele, "1");
String alt = altAllele.getDisplayString();
mWriter.write(alt);
for (int i = 1; i < vc.getAlternateAlleles().size(); i++) {
altAllele = vc.getAlternateAllele(i);
alleleMap.put(altAllele, String.valueOf(i + 1));
alt = altAllele.getDisplayString();
mWriter.write(",");
mWriter.write(alt);
}
} else {
mWriter.write(VCFConstants.EMPTY_ALTERNATE_ALLELE_FIELD);
}
mWriter.write(VCFConstants.FIELD_SEPARATOR);
// QUAL
if (!vc.hasNegLog10PError()) mWriter.write(VCFConstants.MISSING_VALUE_v4);
else mWriter.write(getQualValue(vc.getPhredScaledQual()));
mWriter.write(VCFConstants.FIELD_SEPARATOR);
// FILTER
String filters =
vc.isFiltered()
? ParsingUtils.join(";", ParsingUtils.sortList(vc.getFilters()))
: (filtersWereAppliedToContext || vc.filtersWereApplied()
? VCFConstants.PASSES_FILTERS_v4
: VCFConstants.UNFILTERED);
mWriter.write(filters);
mWriter.write(VCFConstants.FIELD_SEPARATOR);
// INFO
Map<String, String> infoFields = new TreeMap<String, String>();
for (Map.Entry<String, Object> field : vc.getAttributes().entrySet()) {
String key = field.getKey();
if (key.equals(VariantContext.ID_KEY)
|| key.equals(VariantContext.REFERENCE_BASE_FOR_INDEL_KEY)
|| key.equals(VariantContext.UNPARSED_GENOTYPE_MAP_KEY)
|| key.equals(VariantContext.UNPARSED_GENOTYPE_PARSER_KEY)) continue;
String outputValue = formatVCFField(field.getValue());
if (outputValue != null) infoFields.put(key, outputValue);
}
writeInfoString(infoFields);
// FORMAT
if (vc.hasAttribute(VariantContext.UNPARSED_GENOTYPE_MAP_KEY)) {
mWriter.write(VCFConstants.FIELD_SEPARATOR);
mWriter.write(vc.getAttributeAsString(VariantContext.UNPARSED_GENOTYPE_MAP_KEY, ""));
} else {
List<String> genotypeAttributeKeys = new ArrayList<String>();
if (vc.hasGenotypes()) {
genotypeAttributeKeys.addAll(calcVCFGenotypeKeys(vc));
} else if (mHeader.hasGenotypingData()) {
// this needs to be done in case all samples are no-calls
genotypeAttributeKeys.add(VCFConstants.GENOTYPE_KEY);
}
if (genotypeAttributeKeys.size() > 0) {
String genotypeFormatString =
ParsingUtils.join(VCFConstants.GENOTYPE_FIELD_SEPARATOR, genotypeAttributeKeys);
mWriter.write(VCFConstants.FIELD_SEPARATOR);
mWriter.write(genotypeFormatString);
addGenotypeData(vc, alleleMap, genotypeAttributeKeys);
}
}
mWriter.write("\n");
mWriter.flush(); // necessary so that writing to an output stream will work
} catch (IOException e) {
throw new RuntimeException("Unable to write the VCF object to " + locationString());
}
}
/**
* Merges VariantContexts into a single hybrid. Takes genotypes for common samples in priority
* order, if provided. If uniqifySamples is true, the priority order is ignored and names are
* created by concatenating the VC name with the sample name
*
* @param genomeLocParser loc parser
* @param unsortedVCs collection of unsorted VCs
* @param priorityListOfVCs priority list detailing the order in which we should grab the VCs
* @param filteredRecordMergeType merge type for filtered records
* @param genotypeMergeOptions merge option for genotypes
* @param annotateOrigin should we annotate the set it came from?
* @param printMessages should we print messages?
* @param setKey the key name of the set
* @param filteredAreUncalled are filtered records uncalled?
* @param mergeInfoWithMaxAC should we merge in info from the VC with maximum allele count?
* @return new VariantContext representing the merge of unsortedVCs
*/
public static VariantContext simpleMerge(
final GenomeLocParser genomeLocParser,
final Collection<VariantContext> unsortedVCs,
final List<String> priorityListOfVCs,
final FilteredRecordMergeType filteredRecordMergeType,
final GenotypeMergeType genotypeMergeOptions,
final boolean annotateOrigin,
final boolean printMessages,
final String setKey,
final boolean filteredAreUncalled,
final boolean mergeInfoWithMaxAC) {
if (unsortedVCs == null || unsortedVCs.size() == 0) return null;
if (annotateOrigin && priorityListOfVCs == null)
throw new IllegalArgumentException(
"Cannot merge calls and annotate their origins without a complete priority list of VariantContexts");
if (genotypeMergeOptions == GenotypeMergeType.REQUIRE_UNIQUE)
verifyUniqueSampleNames(unsortedVCs);
List<VariantContext> prepaddedVCs =
sortVariantContextsByPriority(unsortedVCs, priorityListOfVCs, genotypeMergeOptions);
// Make sure all variant contexts are padded with reference base in case of indels if necessary
List<VariantContext> VCs = new ArrayList<VariantContext>();
for (VariantContext vc : prepaddedVCs) {
// also a reasonable place to remove filtered calls, if needed
if (!filteredAreUncalled || vc.isNotFiltered())
VCs.add(createVariantContextWithPaddedAlleles(vc, false));
}
if (VCs.size() == 0) // everything is filtered out and we're filteredAreUncalled
return null;
// establish the baseline info from the first VC
final VariantContext first = VCs.get(0);
final String name = first.getSource();
final Allele refAllele = determineReferenceAllele(VCs);
final Set<Allele> alleles = new LinkedHashSet<Allele>();
final Set<String> filters = new TreeSet<String>();
final Map<String, Object> attributes = new TreeMap<String, Object>();
final Set<String> inconsistentAttributes = new HashSet<String>();
final Set<String> variantSources =
new HashSet<
String>(); // contains the set of sources we found in our set of VCs that are variant
final Set<String> rsIDs = new LinkedHashSet<String>(1); // most of the time there's one id
GenomeLoc loc = getLocation(genomeLocParser, first);
int depth = 0;
int maxAC = -1;
final Map<String, Object> attributesWithMaxAC = new TreeMap<String, Object>();
double log10PError = 1;
VariantContext vcWithMaxAC = null;
GenotypesContext genotypes = GenotypesContext.create();
// counting the number of filtered and variant VCs
int nFiltered = 0;
boolean remapped = false;
// cycle through and add info from the other VCs, making sure the loc/reference matches
for (VariantContext vc : VCs) {
if (loc.getStart() != vc.getStart()) // || !first.getReference().equals(vc.getReference()) )
throw new ReviewedStingException(
"BUG: attempting to merge VariantContexts with different start sites: first="
+ first.toString()
+ " second="
+ vc.toString());
if (getLocation(genomeLocParser, vc).size() > loc.size())
loc = getLocation(genomeLocParser, vc); // get the longest location
nFiltered += vc.isFiltered() ? 1 : 0;
if (vc.isVariant()) variantSources.add(vc.getSource());
AlleleMapper alleleMapping = resolveIncompatibleAlleles(refAllele, vc, alleles);
remapped = remapped || alleleMapping.needsRemapping();
alleles.addAll(alleleMapping.values());
mergeGenotypes(
genotypes, vc, alleleMapping, genotypeMergeOptions == GenotypeMergeType.UNIQUIFY);
log10PError = Math.min(log10PError, vc.isVariant() ? vc.getLog10PError() : 1);
filters.addAll(vc.getFilters());
//
// add attributes
//
// special case DP (add it up) and ID (just preserve it)
//
if (vc.hasAttribute(VCFConstants.DEPTH_KEY))
depth += vc.getAttributeAsInt(VCFConstants.DEPTH_KEY, 0);
if (vc.hasID()) rsIDs.add(vc.getID());
if (mergeInfoWithMaxAC && vc.hasAttribute(VCFConstants.ALLELE_COUNT_KEY)) {
String rawAlleleCounts = vc.getAttributeAsString(VCFConstants.ALLELE_COUNT_KEY, null);
// lets see if the string contains a , separator
if (rawAlleleCounts.contains(VCFConstants.INFO_FIELD_ARRAY_SEPARATOR)) {
List<String> alleleCountArray =
Arrays.asList(
rawAlleleCounts
.substring(1, rawAlleleCounts.length() - 1)
.split(VCFConstants.INFO_FIELD_ARRAY_SEPARATOR));
for (String alleleCount : alleleCountArray) {
final int ac = Integer.valueOf(alleleCount.trim());
if (ac > maxAC) {
maxAC = ac;
vcWithMaxAC = vc;
}
}
} else {
final int ac = Integer.valueOf(rawAlleleCounts);
if (ac > maxAC) {
maxAC = ac;
vcWithMaxAC = vc;
}
}
}
for (Map.Entry<String, Object> p : vc.getAttributes().entrySet()) {
String key = p.getKey();
// if we don't like the key already, don't go anywhere
if (!inconsistentAttributes.contains(key)) {
boolean alreadyFound = attributes.containsKey(key);
Object boundValue = attributes.get(key);
boolean boundIsMissingValue =
alreadyFound && boundValue.equals(VCFConstants.MISSING_VALUE_v4);
if (alreadyFound && !boundValue.equals(p.getValue()) && !boundIsMissingValue) {
// we found the value but we're inconsistent, put it in the exclude list
// System.out.printf("Inconsistent INFO values: %s => %s and %s%n", key, boundValue,
// p.getValue());
inconsistentAttributes.add(key);
attributes.remove(key);
} else if (!alreadyFound || boundIsMissingValue) { // no value
// if ( vc != first ) System.out.printf("Adding key %s => %s%n", p.getKey(),
// p.getValue());
attributes.put(key, p.getValue());
}
}
}
}
// if we have more alternate alleles in the merged VC than in one or more of the
// original VCs, we need to strip out the GL/PLs (because they are no longer accurate), as well
// as allele-dependent attributes like AC,AF
for (VariantContext vc : VCs) {
if (vc.alleles.size() == 1) continue;
if (hasPLIncompatibleAlleles(alleles, vc.alleles)) {
if (!genotypes.isEmpty())
logger.warn(
String.format(
"Stripping PLs at %s due incompatible alleles merged=%s vs. single=%s",
genomeLocParser.createGenomeLoc(vc), alleles, vc.alleles));
genotypes = stripPLs(genotypes);
// this will remove stale AC,AF attributed from vc
calculateChromosomeCounts(vc, attributes, true);
break;
}
}
// take the VC with the maxAC and pull the attributes into a modifiable map
if (mergeInfoWithMaxAC && vcWithMaxAC != null) {
attributesWithMaxAC.putAll(vcWithMaxAC.getAttributes());
}
// if at least one record was unfiltered and we want a union, clear all of the filters
if ((filteredRecordMergeType == FilteredRecordMergeType.KEEP_IF_ANY_UNFILTERED
&& nFiltered != VCs.size())
|| filteredRecordMergeType == FilteredRecordMergeType.KEEP_UNCONDITIONAL) filters.clear();
if (annotateOrigin) { // we care about where the call came from
String setValue;
if (nFiltered == 0
&& variantSources.size() == priorityListOfVCs.size()) // nothing was unfiltered
setValue = MERGE_INTERSECTION;
else if (nFiltered == VCs.size()) // everything was filtered out
setValue = MERGE_FILTER_IN_ALL;
else if (variantSources.isEmpty()) // everyone was reference
setValue = MERGE_REF_IN_ALL;
else {
LinkedHashSet<String> s = new LinkedHashSet<String>();
for (VariantContext vc : VCs)
if (vc.isVariant())
s.add(vc.isFiltered() ? MERGE_FILTER_PREFIX + vc.getSource() : vc.getSource());
setValue = Utils.join("-", s);
}
if (setKey != null) {
attributes.put(setKey, setValue);
if (mergeInfoWithMaxAC && vcWithMaxAC != null) {
attributesWithMaxAC.put(setKey, vcWithMaxAC.getSource());
}
}
}
if (depth > 0) attributes.put(VCFConstants.DEPTH_KEY, String.valueOf(depth));
final String ID = rsIDs.isEmpty() ? VCFConstants.EMPTY_ID_FIELD : Utils.join(",", rsIDs);
final VariantContextBuilder builder = new VariantContextBuilder().source(name).id(ID);
builder.loc(loc.getContig(), loc.getStart(), loc.getStop());
builder.alleles(alleles);
builder.genotypes(genotypes);
builder.log10PError(log10PError);
builder.filters(filters).attributes(mergeInfoWithMaxAC ? attributesWithMaxAC : attributes);
// Trim the padded bases of all alleles if necessary
VariantContext merged = createVariantContextWithTrimmedAlleles(builder.make());
if (printMessages && remapped) System.out.printf("Remapped => %s%n", merged);
return merged;
}