Examples with GenotypeBuilder htsjdk.variant.variantcontext.GenotypeBuilder used on opensource projects

Example 16 with GenotypeBuilder

use of htsjdk.variant.variantcontext.GenotypeBuilder in project gatk by broadinstitute.

the class DepthPerAlleleBySample method annotate.

@Override
public void annotate(final ReferenceContext ref, final VariantContext vc, final Genotype g, final GenotypeBuilder gb, final ReadLikelihoods<Allele> likelihoods) {
    Utils.nonNull(gb, "gb is null");
    Utils.nonNull(vc, "vc is null");
    if (g == null || !g.isCalled() || likelihoods == null) {
        return;
    }
    final Set<Allele> alleles = new LinkedHashSet<>(vc.getAlleles());
    // make sure that there's a meaningful relationship between the alleles in the likelihoods and our VariantContext
    Utils.validateArg(likelihoods.alleles().containsAll(alleles), () -> "VC alleles " + alleles + " not a  subset of ReadLikelihoods alleles " + likelihoods.alleles());
    final Map<Allele, Integer> alleleCounts = new LinkedHashMap<>();
    for (final Allele allele : vc.getAlleles()) {
        alleleCounts.put(allele, 0);
    }
    final Map<Allele, List<Allele>> alleleSubset = alleles.stream().collect(Collectors.toMap(a -> a, Arrays::asList));
    final ReadLikelihoods<Allele> subsettedLikelihoods = likelihoods.marginalize(alleleSubset);
    subsettedLikelihoods.bestAlleles(g.getSampleName()).stream().filter(ba -> ba.isInformative()).forEach(ba -> alleleCounts.compute(ba.allele, (allele, prevCount) -> prevCount + 1));
    final int[] counts = new int[alleleCounts.size()];
    //first one in AD is always ref
    counts[0] = alleleCounts.get(vc.getReference());
    for (int i = 0; i < vc.getAlternateAlleles().size(); i++) {
        counts[i + 1] = alleleCounts.get(vc.getAlternateAllele(i));
    }
    gb.AD(counts);
}

Example 17 with GenotypeBuilder

use of htsjdk.variant.variantcontext.GenotypeBuilder in project gatk-protected by broadinstitute.

the class StrandArtifact method annotate.

@Override
public void annotate(final ReferenceContext ref, final VariantContext vc, final Genotype g, final GenotypeBuilder gb, final ReadLikelihoods<Allele> likelihoods) {
    Utils.nonNull(gb);
    Utils.nonNull(vc);
    Utils.nonNull(likelihoods);
    // do not annotate the genotype fields for normal
    if (g.isHomRef()) {
        return;
    }
    pi.put(NO_ARTIFACT, 0.95);
    pi.put(ART_FWD, 0.025);
    pi.put(ART_REV, 0.025);
    // We use the allele with highest LOD score
    final double[] tumorLods = GATKProtectedVariantContextUtils.getAttributeAsDoubleArray(vc, GATKVCFConstants.TUMOR_LOD_KEY, () -> null, -1);
    final int indexOfMaxTumorLod = MathUtils.maxElementIndex(tumorLods);
    final Allele altAlelle = vc.getAlternateAllele(indexOfMaxTumorLod);
    final Collection<ReadLikelihoods<Allele>.BestAllele<Allele>> bestAlleles = likelihoods.bestAlleles(g.getSampleName());
    final int numFwdAltReads = (int) bestAlleles.stream().filter(ba -> !ba.read.isReverseStrand() && ba.isInformative() && ba.allele.equals(altAlelle)).count();
    final int numRevAltReads = (int) bestAlleles.stream().filter(ba -> ba.read.isReverseStrand() && ba.isInformative() && ba.allele.equals(altAlelle)).count();
    final int numFwdReads = (int) bestAlleles.stream().filter(ba -> !ba.read.isReverseStrand() && ba.isInformative()).count();
    final int numRevReads = (int) bestAlleles.stream().filter(ba -> ba.read.isReverseStrand() && ba.isInformative()).count();
    final int numAltReads = numFwdAltReads + numRevAltReads;
    final int numReads = numFwdReads + numRevReads;
    final EnumMap<StrandArtifactZ, Double> unnormalized_posterior_probabilities = new EnumMap<>(StrandArtifactZ.class);
    final EnumMap<StrandArtifactZ, Double> maximum_a_posteriori_allele_fraction_estimates = new EnumMap<>(StrandArtifactZ.class);
    /*** Compute the posterior probability of ARTIFACT_FWD and ARTIFACT_REV; it's a double integral over f and epsilon ***/
    // the integrand is a polynomial of degree n, where n is the number of reads at the locus
    // thus to integrate exactly with Gauss-Legendre we need (n/2)+1 points
    final int numIntegPointsForAlleleFraction = numReads / 2 + 1;
    final int numIntegPointsForEpsilon = (numReads + ALPHA + BETA - 2) / 2 + 1;
    final double likelihoodForArtifactFwd = IntegrationUtils.integrate2d((f, epsilon) -> getIntegrandGivenArtifact(f, epsilon, numFwdReads, numRevReads, numFwdAltReads, numRevAltReads), 0.0, 1.0, numIntegPointsForAlleleFraction, 0.0, 1.0, numIntegPointsForEpsilon);
    final double likelihoodForArtifactRev = IntegrationUtils.integrate2d((f, epsilon) -> getIntegrandGivenArtifact(f, epsilon, numRevReads, numFwdReads, numRevAltReads, numFwdAltReads), 0.0, 1.0, numIntegPointsForAlleleFraction, 0.0, 1.0, numIntegPointsForEpsilon);
    unnormalized_posterior_probabilities.put(ART_FWD, pi.get(ART_FWD) * likelihoodForArtifactFwd);
    unnormalized_posterior_probabilities.put(ART_REV, pi.get(ART_REV) * likelihoodForArtifactRev);
    /*** Compute the posterior probability of NO_ARTIFACT; evaluate a single integral over the allele fraction ***/
    final double likelihoodForNoArtifact = IntegrationUtils.integrate(f -> getIntegrandGivenNoArtifact(f, numFwdReads, numRevReads, numFwdAltReads, numRevAltReads), 0.0, 1.0, numIntegPointsForAlleleFraction);
    unnormalized_posterior_probabilities.put(NO_ARTIFACT, pi.get(NO_ARTIFACT) * likelihoodForNoArtifact);
    final double[] posterior_probabilities = MathUtils.normalizeFromRealSpace(unnormalized_posterior_probabilities.values().stream().mapToDouble(Double::doubleValue).toArray());
    /*** Compute the maximum a posteriori estimate for allele fraction given strand artifact ***/
    // For a fixed f, integrate the double integral over epsilons. This gives us the likelihood p(x^+, x^- | f, z) for a fixed f, which is proportional to
    // the posterior probability of p(f | x^+, x^-, z)
    final int numSamplePoints = 100;
    final double[] samplePoints = GATKProtectedMathUtils.createEvenlySpacedPoints(0.0, 1.0, numSamplePoints);
    double[] likelihoodsGivenForwardArtifact = new double[numSamplePoints];
    double[] likelihoodsGivenReverseArtifact = new double[numSamplePoints];
    for (int i = 0; i < samplePoints.length; i++) {
        final double f = samplePoints[i];
        likelihoodsGivenForwardArtifact[i] = IntegrationUtils.integrate(epsilon -> getIntegrandGivenArtifact(f, epsilon, numFwdReads, numRevReads, numFwdAltReads, numRevAltReads), 0.0, 1.0, numIntegPointsForEpsilon);
        likelihoodsGivenReverseArtifact[i] = IntegrationUtils.integrate(epsilon -> getIntegrandGivenArtifact(f, epsilon, numRevReads, numFwdReads, numRevAltReads, numFwdAltReads), 0.0, 1.0, numIntegPointsForEpsilon);
    }
    final int maxAlleleFractionIndexFwd = MathUtils.maxElementIndex(likelihoodsGivenForwardArtifact);
    final int maxAlleleFractionIndexRev = MathUtils.maxElementIndex(likelihoodsGivenReverseArtifact);
    maximum_a_posteriori_allele_fraction_estimates.put(ART_FWD, samplePoints[maxAlleleFractionIndexFwd]);
    maximum_a_posteriori_allele_fraction_estimates.put(ART_REV, samplePoints[maxAlleleFractionIndexRev]);
    // In the absence of strand artifact, MAP estimate for f reduces to the sample alt allele fraction
    maximum_a_posteriori_allele_fraction_estimates.put(NO_ARTIFACT, (double) numAltReads / numReads);
    gb.attribute(POSTERIOR_PROBABILITIES_KEY, posterior_probabilities);
    gb.attribute(MAP_ALLELE_FRACTIONS_KEY, maximum_a_posteriori_allele_fraction_estimates.values().stream().mapToDouble(Double::doubleValue).toArray());
}