Struct libprosic::model::priors::tumor_normal::TumorNormalModel [−] [src]

pub struct TumorNormalModel {
    pub normal_model: InfiniteSitesNeutralVariationModel,
    pub allele_freqs_tumor: ContinuousAlleleFreqs,
    pub grid_points: usize,
    // some fields omitted
}

Tumor-normal prior model using ploidy, heterozygosity (in normal tissue) and tumor mutation rate per effective cell division. The latter is the quotient mu/beta, with mu being the mutation rate and beta being the fraction of effective cell divisions (both lineages survive). Alone, the parameters are not observable. However, mu/beta can be estimated from e.g. SNV calls. It is the slope of the linear model y = mu/beta * (x - 1 / fmax), withx` being the reciprocal of the observed allele frequencies and y being the number of observed mutations corresponding to each frequency (see Williams et al. Nature Genetics 2016).

Based on the Williams model, the tail probability of a somatic allele frequency F > f can be expressed as Pr(F > f) = M(f) / n = mu/beta (1 / f - 1 / fmax) / n with n being the size of the genome and fmax is the expected allele frequency of clonal variants at the beginning of tumor evolution. From this, we can obtain the cumulative distribution function as Pr(F <= f) = 1 - Pr(F > f). Consequently, the density becomes the first derivative, i.e. Pr(F = f) = - M(f)' / n = mu/beta * 1/n * 1/f² for f>=fmin with fmin = sqrt(mu/beta * 1/n).

The prior probability for a germline allele frequency f (e.g. 0.0, 0.5, 1.0) in the tumor is calculated with an InfiniteSitesNeutralVariationModel. This is valid since clonal variants come from the underlying normal tissue and Williams model assumes that allele frequencies do not change during tumor evolution (no genetic drift, no selection).

For the final prior, we consider a given tumor purity and calculate the combined prior for all possible allele frequency combinations satisfying af = purity * af_tumor + (1-purity) * af_normal.

Fields

normal_model: InfiniteSitesNeutralVariationModel allele_freqs_tumor: ContinuousAlleleFreqs grid_points: usize

Methods

impl TumorNormalModel
[src]

fn new(ploidy: u32, effective_mutation_rate: f64, deletion_factor: f64, insertion_factor: f64, genome_size: u64, heterozygosity: Prob) -> Self

Create new model.

Arguments

  • ploidy - the ploidy in the corresponding normal sample (e.g. 2 for diploid)
  • effective_mutation_rate - the SNV mutation rate per effective cell division in the tumor
  • deletion_factor - ratio of deletions compared to SNV mutation rate
  • insertion_factor - ratio of insertions compared to SNV mutation rate
  • genome_size - the size of the genome
  • heterozygosity - expected heterozygosity in the corresponding normal

fn somatic_prior_prob(&self, af_somatic: AlleleFreq, variant: Variant) -> LogProb

fn normal_prior_prob(&self, af_normal: AlleleFreq, variant: Variant) -> LogProb

Trait Implementations

impl PairModel<ContinuousAlleleFreqs, DiscreteAlleleFreqs> for TumorNormalModel
[src]

fn prior_prob(&self, af_tumor: AlleleFreq, af_normal: AlleleFreq, variant: Variant) -> LogProb

Calculate prior probability of given combination of allele frequencies.

fn joint_prob<L, O>(&self, af_tumor: &ContinuousAlleleFreqs, af_normal: &DiscreteAlleleFreqs, likelihood_tumor: &L, likelihood_normal: &O, variant: Variant) -> LogProb where L: Fn(AlleleFreq, AlleleFreq) -> LogProb, O: Fn(AlleleFreq, AlleleFreq) -> LogProb

Calculate joint probability of prior with likelihoods for given allele frequency ranges.

fn marginal_prob<L, O>(&self, likelihood_tumor: &L, likelihood_normal: &O, variant: Variant) -> LogProb where L: Fn(AlleleFreq, AlleleFreq) -> LogProb, O: Fn(AlleleFreq, AlleleFreq) -> LogProb

Calculate marginal probability.

fn map<L, O>(&self, likelihood_tumor: &L, likelihood_normal: &O, variant: Variant) -> (AlleleFreq, AlleleFreq) where L: Fn(AlleleFreq, AlleleFreq) -> LogProb, O: Fn(AlleleFreq, AlleleFreq) -> LogProb

Calculate maximum a posteriori probability estimate of allele frequencies.

fn allele_freqs(&self) -> (&ContinuousAlleleFreqs, &DiscreteAlleleFreqs)

Return allele frequency spectra.