joint caller parameters
the allele under consideration
metadata for each sample: is it tumor / normal, is it DNA / RNA
evidence for the pooled normal DNA samples
evidence for the pooled tumor DNA samples
evidences for each sample individually, corresponding to inputs
There are times when we want to treat all the per-sample evidences and pooled evidences together.
There are times when we want to treat all the per-sample evidences and pooled evidences together. We do this
with a sequence called allEvidences
that has the evidence for each sample followed by that of the two pooled
samples. The normal dna pooled sample is found at index normalDNAPooledIndex
and the tumor pooled is at index
tumorDNAPooledIndex
in allEvidences
.
the allele under consideration
Return a new instance with all available annotations computed.
Names of filters that are causing this potential call to be considered filtered.
Names of filters that are causing this potential call to be considered filtered.
If this function returns an empty set, then the call is not filtered. That could be because it is a valid passing variant (in which isCalled will be true) or because it was force called and none of the filters happened to fail.
Some filters apply to individual samples (e.g. strand bias), whereas others apply to all samples for an allele at a site (e.g. insufficient evidence in normal). The former are implemented in SampleAlleleEvidenceAnnotation, and the latter are in CallsAtSiteAnnotation.
The handling for the first kind of filter (per-sample filters) is a bit subtle here, because we do not care about filters that are failing on some samples when, even if those samples were ignored, a call would still be triggered. For example, if two tumors are triggering a somatic variant call and one of them fails the strand bias filter, we do *not* want to treat this call as filtered. If however both of those samples fail the strand bias filter, or even if they both fail different filters, then we do want to consider this call as filtered. A consequence of this is that if isCalled is false, then this function will never return any individual sample filters.
true if this potential call should be considered to be filtered
Called germline genotype.
Called germline genotype. We currently just use the maximum posterior from the pooled normal data.
The individual TumorDNASampleAlleleEvidence and NormalDNASampleAlleleEvidence instances calculate the mixture likelihoods, but do not know about the prior.
The individual TumorDNASampleAlleleEvidence and NormalDNASampleAlleleEvidence instances calculate the mixture likelihoods, but do not know about the prior. We establish the germline prior here.
The rationale for having this in the current class instead of the SampleAlleleEvidence classes is that eventually we are going to use RNA evidence and phasing information to influence the posteriors, and we'll need to calculate that here, since the individual evidence classes only know about a single sample.
Note that at no point are we working with normalized priors, likelihoods, or posteriors. All of these are specified just up to proportionality and do not need to sum to 1.
pair of alleles to return prior for
negative log-base 10 prior probability for that allele. Since log probs are negative, this number will be *positive* (it's the negative of a negative).
metadata for each sample: is it tumor / normal, is it DNA / RNA
Are we making a germline or somatic call?
Are we making a germline call here?
Are we making a somatic call here?
evidence for the pooled normal DNA samples
joint caller parameters
Log10 posterior probabilities for each possible germline genotype in each normal sample.
Log10 posterior probabilities for each possible germline genotype in each normal sample.
Map is from input index (i.e. index into inputs and sampleEvidences) to mixture posteriors.
The posteriors are plain log10 logprobs. They are negative. The maximum a posteriori estimate is the greatest (i.e. least negative, closest to 0) posterior.
Log10 posterior probabilities for a somatic variant in each tumor DNA sample.
Log10 posterior probabilities for a somatic variant in each tumor DNA sample. See perNormalSampleGermlinePosteriors.
Map {input index -> {{Allele -> Frequency} -> Posterior probability}
Maximum a posteriori somatic mixtures for each tumor sample.
Log10 posterior probabilities for a somatic variant in each tumor RNA sample.
Maximum a posteriori somatic mixtures for each tumor sample.
evidences for each sample individually, corresponding to inputs
evidence for the pooled tumor DNA samples
Indices of tumor samples that triggered a call.
Indices of tumor rna samples with expression
Summarizes the evidence for a single allele across any number of samples.
Keeps track of the evidence for the allele in each sample individually, and also the pooled normal and tumor DNA.
joint caller parameters
the allele under consideration
metadata for each sample: is it tumor / normal, is it DNA / RNA
evidence for the pooled normal DNA samples
evidence for the pooled tumor DNA samples
evidences for each sample individually, corresponding to inputs