private boolean isConcordant(VariantContext vc, Collection<VariantContext> compVCs) {
if (vc == null || compVCs == null || compVCs.isEmpty()) return false;
// if we're not looking for specific samples then the fact that we have both VCs is enough to
// call it concordant.
if (NO_SAMPLES_SPECIFIED) return true;
// make a list of all samples contained in this variant VC that are being tracked by the user
// command line arguments.
Set<String> variantSamples = vc.getSampleNames();
variantSamples.retainAll(samples);
// check if we can find all samples from the variant rod in the comp rod.
for (String sample : variantSamples) {
boolean foundSample = false;
for (VariantContext compVC : compVCs) {
Genotype varG = vc.getGenotype(sample);
Genotype compG = compVC.getGenotype(sample);
if (haveSameGenotypes(varG, compG)) {
foundSample = true;
break;
}
}
// if at least one sample doesn't have the same genotype, we don't have concordance
if (!foundSample) {
return false;
}
}
return true;
}
public int getNumberOfSamplesForEvaluation() {
if (sampleNamesForEvaluation != null && !sampleNamesForEvaluation.isEmpty())
return sampleNamesForEvaluation.size();
else {
return numSamplesFromArgument;
}
}
private void initializeVcfWriter() {
final List<String> inputNames = Arrays.asList(validation.getName());
// setup the header fields
Set<VCFHeaderLine> hInfo = new HashSet<VCFHeaderLine>();
hInfo.addAll(VCFUtils.getHeaderFields(getToolkit(), inputNames));
hInfo.add(
new VCFFilterHeaderLine(
"bootstrap",
"This site used for genotype bootstrapping with ProduceBeagleInputWalker"));
bootstrapVCFOutput.writeHeader(
new VCFHeader(hInfo, SampleUtils.getUniqueSamplesFromRods(getToolkit(), inputNames)));
}
/**
* Helper method to subset a VC record, modifying some metadata stored in the INFO field (i.e. AN,
* AC, AF).
*
* @param vc the VariantContext record to subset
* @param samples the samples to extract
* @return the subsetted VariantContext
*/
private VariantContext subsetRecord(VariantContext vc, Set<String> samples) {
if (samples == null || samples.isEmpty()) return vc;
ArrayList<Genotype> genotypes = new ArrayList<Genotype>();
for (Map.Entry<String, Genotype> genotypePair : vc.getGenotypes().entrySet()) {
if (samples.contains(genotypePair.getKey())) genotypes.add(genotypePair.getValue());
}
VariantContext sub = vc.subContextFromGenotypes(genotypes, vc.getAlleles());
// if we have fewer alternate alleles in the selected VC than in the original VC, we need to
// strip out the GL/PLs (because they are no longer accurate)
if (vc.getAlleles().size() != sub.getAlleles().size())
sub = VariantContext.modifyGenotypes(sub, VariantContextUtils.stripPLs(vc.getGenotypes()));
HashMap<String, Object> attributes = new HashMap<String, Object>(sub.getAttributes());
int depth = 0;
for (String sample : sub.getSampleNames()) {
Genotype g = sub.getGenotype(sample);
if (g.isNotFiltered() && g.isCalled()) {
String dp = (String) g.getAttribute("DP");
if (dp != null
&& !dp.equals(VCFConstants.MISSING_DEPTH_v3)
&& !dp.equals(VCFConstants.MISSING_VALUE_v4)) {
depth += Integer.valueOf(dp);
}
}
}
if (KEEP_ORIGINAL_CHR_COUNTS) {
if (attributes.containsKey(VCFConstants.ALLELE_COUNT_KEY))
attributes.put("AC_Orig", attributes.get(VCFConstants.ALLELE_COUNT_KEY));
if (attributes.containsKey(VCFConstants.ALLELE_FREQUENCY_KEY))
attributes.put("AF_Orig", attributes.get(VCFConstants.ALLELE_FREQUENCY_KEY));
if (attributes.containsKey(VCFConstants.ALLELE_NUMBER_KEY))
attributes.put("AN_Orig", attributes.get(VCFConstants.ALLELE_NUMBER_KEY));
}
VariantContextUtils.calculateChromosomeCounts(sub, attributes, false);
attributes.put("DP", depth);
sub = VariantContext.modifyAttributes(sub, attributes);
return sub;
}
/**
* Compares the covariate report lists.
*
* @param diffs map where to annotate the difference.
* @param other the argument collection to compare against.
* @param thisRole the name for this argument collection that makes sense to the user.
* @param otherRole the name for the other argument collection that makes sense to the end user.
* @return <code>true</code> if a difference was found.
*/
@Requires("diffs != null && other != null && thisRole != null && otherRole != null")
private boolean compareRequestedCovariates(
final Map<String, String> diffs,
final RecalibrationArgumentCollection other,
final String thisRole,
final String otherRole) {
final Set<String> beforeNames = new HashSet<>(this.COVARIATES.length);
final Set<String> afterNames = new HashSet<>(other.COVARIATES.length);
Utils.addAll(beforeNames, this.COVARIATES);
Utils.addAll(afterNames, other.COVARIATES);
final Set<String> intersect = new HashSet<>(Math.min(beforeNames.size(), afterNames.size()));
intersect.addAll(beforeNames);
intersect.retainAll(afterNames);
String diffMessage = null;
if (intersect.size()
== 0) { // In practice this is not possible due to required covariates but...
diffMessage =
String.format(
"There are no common covariates between '%s' and '%s'"
+ " recalibrator reports. Covariates in '%s': {%s}. Covariates in '%s': {%s}.",
thisRole,
otherRole,
thisRole,
Utils.join(", ", this.COVARIATES),
otherRole,
Utils.join(",", other.COVARIATES));
} else if (intersect.size() != beforeNames.size() || intersect.size() != afterNames.size()) {
beforeNames.removeAll(intersect);
afterNames.removeAll(intersect);
diffMessage =
String.format(
"There are differences in the set of covariates requested in the"
+ " '%s' and '%s' recalibrator reports. "
+ " Exclusive to '%s': {%s}. Exclusive to '%s': {%s}.",
thisRole,
otherRole,
thisRole,
Utils.join(", ", beforeNames),
otherRole,
Utils.join(", ", afterNames));
}
if (diffMessage != null) {
diffs.put("covariate", diffMessage);
return true;
} else {
return false;
}
}
/** Initialize the stratifications, evaluations, evaluation contexts, and reporting object */
public void initialize() {
// Just list the modules, and exit quickly.
if (LIST) {
variantEvalUtils.listModulesAndExit();
}
// maintain the full list of comps
comps.addAll(compsProvided);
if (dbsnp.dbsnp.isBound()) {
comps.add(dbsnp.dbsnp);
knowns.add(dbsnp.dbsnp);
}
// Add a dummy comp track if none exists
if (comps.size() == 0)
comps.add(
new RodBinding<VariantContext>(VariantContext.class, "none", "UNBOUND", "", new Tags()));
// Set up set of additional knowns
for (RodBinding<VariantContext> compRod : comps) {
if (KNOWN_NAMES.contains(compRod.getName())) knowns.add(compRod);
}
// Now that we have all the rods categorized, determine the sample list from the eval rods.
Map<String, VCFHeader> vcfRods = VCFUtils.getVCFHeadersFromRods(getToolkit(), evals);
Set<String> vcfSamples =
SampleUtils.getSampleList(vcfRods, VariantContextUtils.GenotypeMergeType.REQUIRE_UNIQUE);
// Load the sample list
sampleNamesForEvaluation.addAll(
SampleUtils.getSamplesFromCommandLineInput(vcfSamples, SAMPLE_EXPRESSIONS));
numSamples = NUM_SAMPLES > 0 ? NUM_SAMPLES : sampleNamesForEvaluation.size();
if (Arrays.asList(STRATIFICATIONS_TO_USE).contains("Sample")) {
sampleNamesForStratification.addAll(sampleNamesForEvaluation);
}
sampleNamesForStratification.add(ALL_SAMPLE_NAME);
// Initialize select expressions
for (VariantContextUtils.JexlVCMatchExp jexl :
VariantContextUtils.initializeMatchExps(SELECT_NAMES, SELECT_EXPS)) {
SortableJexlVCMatchExp sjexl = new SortableJexlVCMatchExp(jexl.name, jexl.exp);
jexlExpressions.add(sjexl);
}
// Initialize the set of stratifications and evaluations to use
stratificationObjects =
variantEvalUtils.initializeStratificationObjects(
this, NO_STANDARD_STRATIFICATIONS, STRATIFICATIONS_TO_USE);
Set<Class<? extends VariantEvaluator>> evaluationObjects =
variantEvalUtils.initializeEvaluationObjects(NO_STANDARD_MODULES, MODULES_TO_USE);
for (VariantStratifier vs : getStratificationObjects()) {
if (vs.getName().equals("Filter")) byFilterIsEnabled = true;
else if (vs.getName().equals("Sample")) perSampleIsEnabled = true;
}
if (intervalsFile != null) {
boolean fail = true;
for (final VariantStratifier vs : stratificationObjects) {
if (vs.getClass().equals(IntervalStratification.class)) fail = false;
}
if (fail)
throw new UserException.BadArgumentValue(
"ST", "stratIntervals argument provided but -ST IntervalStratification not provided");
}
// Initialize the evaluation contexts
evaluationContexts =
variantEvalUtils.initializeEvaluationContexts(
stratificationObjects, evaluationObjects, null, null);
// Initialize report table
report = variantEvalUtils.initializeGATKReport(stratificationObjects, evaluationObjects);
// Load ancestral alignments
if (ancestralAlignmentsFile != null) {
try {
ancestralAlignments = new IndexedFastaSequenceFile(ancestralAlignmentsFile);
} catch (FileNotFoundException e) {
throw new ReviewedStingException(
String.format(
"The ancestral alignments file, '%s', could not be found",
ancestralAlignmentsFile.getAbsolutePath()));
}
}
// initialize CNVs
if (knownCNVsFile != null) {
knownCNVsByContig = createIntervalTreeByContig(knownCNVsFile);
}
}
/** Set up the VCF writer, the sample expressions and regexs, and the JEXL matcher */
public void initialize() {
// Get list of samples to include in the output
List<String> rodNames = Arrays.asList(variantCollection.variants.getName());
Map<String, VCFHeader> vcfRods = VCFUtils.getVCFHeadersFromRods(getToolkit(), rodNames);
TreeSet<String> vcfSamples =
new TreeSet<String>(
SampleUtils.getSampleList(
vcfRods, VariantContextUtils.GenotypeMergeType.REQUIRE_UNIQUE));
Collection<String> samplesFromFile = SampleUtils.getSamplesFromFiles(sampleFiles);
Collection<String> samplesFromExpressions =
SampleUtils.matchSamplesExpressions(vcfSamples, sampleExpressions);
// first, add any requested samples
samples.addAll(samplesFromFile);
samples.addAll(samplesFromExpressions);
samples.addAll(sampleNames);
// if none were requested, we want all of them
if (samples.isEmpty()) {
samples.addAll(vcfSamples);
NO_SAMPLES_SPECIFIED = true;
}
// now, exclude any requested samples
Collection<String> XLsamplesFromFile = SampleUtils.getSamplesFromFiles(XLsampleFiles);
samples.removeAll(XLsamplesFromFile);
samples.removeAll(XLsampleNames);
if (samples.size() == 0 && !NO_SAMPLES_SPECIFIED)
throw new UserException(
"All samples requested to be included were also requested to be excluded.");
for (String sample : samples) logger.info("Including sample '" + sample + "'");
// if user specified types to include, add these, otherwise, add all possible variant context
// types to list of vc types to include
if (TYPES_TO_INCLUDE.isEmpty()) {
for (VariantContext.Type t : VariantContext.Type.values()) selectedTypes.add(t);
} else {
for (VariantContext.Type t : TYPES_TO_INCLUDE) selectedTypes.add(t);
}
// Initialize VCF header
Set<VCFHeaderLine> headerLines = VCFUtils.smartMergeHeaders(vcfRods.values(), logger);
headerLines.add(new VCFHeaderLine("source", "SelectVariants"));
if (KEEP_ORIGINAL_CHR_COUNTS) {
headerLines.add(
new VCFFormatHeaderLine("AC_Orig", 1, VCFHeaderLineType.Integer, "Original AC"));
headerLines.add(
new VCFFormatHeaderLine("AF_Orig", 1, VCFHeaderLineType.Float, "Original AF"));
headerLines.add(
new VCFFormatHeaderLine("AN_Orig", 1, VCFHeaderLineType.Integer, "Original AN"));
}
vcfWriter.writeHeader(new VCFHeader(headerLines, samples));
for (int i = 0; i < SELECT_EXPRESSIONS.size(); i++) {
// It's not necessary that the user supply select names for the JEXL expressions, since those
// expressions will only be needed for omitting records. Make up the select names here.
selectNames.add(String.format("select-%d", i));
}
jexls = VariantContextUtils.initializeMatchExps(selectNames, SELECT_EXPRESSIONS);
// Look at the parameters to decide which analysis to perform
DISCORDANCE_ONLY = discordanceTrack.isBound();
if (DISCORDANCE_ONLY)
logger.info(
"Selecting only variants discordant with the track: " + discordanceTrack.getName());
CONCORDANCE_ONLY = concordanceTrack.isBound();
if (CONCORDANCE_ONLY)
logger.info(
"Selecting only variants concordant with the track: " + concordanceTrack.getName());
if (MENDELIAN_VIOLATIONS) {
if (FAMILY_STRUCTURE_FILE != null) {
try {
for (final String line : new XReadLines(FAMILY_STRUCTURE_FILE)) {
MendelianViolation mv =
new MendelianViolation(line, MENDELIAN_VIOLATION_QUAL_THRESHOLD);
if (samples.contains(mv.getSampleChild())
&& samples.contains(mv.getSampleDad())
&& samples.contains(mv.getSampleMom())) mvSet.add(mv);
}
} catch (FileNotFoundException e) {
throw new UserException.CouldNotReadInputFile(FAMILY_STRUCTURE_FILE, e);
}
if (outMVFile != null)
try {
outMVFileStream = new PrintStream(outMVFile);
} catch (FileNotFoundException e) {
throw new UserException.CouldNotCreateOutputFile(
outMVFile, "Can't open output file", e);
}
} else
mvSet.add(new MendelianViolation(FAMILY_STRUCTURE, MENDELIAN_VIOLATION_QUAL_THRESHOLD));
} else if (!FAMILY_STRUCTURE.isEmpty()) {
mvSet.add(new MendelianViolation(FAMILY_STRUCTURE, MENDELIAN_VIOLATION_QUAL_THRESHOLD));
MENDELIAN_VIOLATIONS = true;
}
SELECT_RANDOM_NUMBER = numRandom > 0;
if (SELECT_RANDOM_NUMBER) {
logger.info("Selecting " + numRandom + " variants at random from the variant track");
variantArray = new RandomVariantStructure[numRandom];
}
SELECT_RANDOM_FRACTION = fractionRandom > 0;
if (SELECT_RANDOM_FRACTION)
logger.info(
"Selecting approximately "
+ 100.0 * fractionRandom
+ "% of the variants at random from the variant track");
}