CalcParentProbs: Calculate assignment probabilities

In sequoia: Pedigree Inference from SNPs

Calculate assignment probabilities

Description

For each assigned offspring-parent pair, calculate the probability they are parent-offspring vs otherwise related. Probabilities are scaled to sum to one across all possible* relationships between the pair or trio; see Details.

Usage

CalcParentProbs(Pedigree = NULL, GenoM = NULL, quiet = FALSE, nCores = 1, ...)

Arguments

Pedigree	dataframe with columns id-dam-sire. By default, any non-genotyped individuals are 'dummified'; use Module='par' to ignore them.
GenoM	numeric matrix with genotype data: One row per individual, one column per SNP, coded as 0, 1, 2, missing values as a negative number or NA. You can reformat data with GenoConvert, or use other packages to get it into a genlight object and then use as.matrix.
quiet	logical, suppress messages. No progress is printed when >1 core is used.
nCores	number of computer cores to use. If 2 or 4, package parallel is used (other values are not applicable).
...	Additional arguments passed to CalcPairLL, such as the genotyping error rate Err, age information in LifeHistData and AgePrior, or InclDup to include the probability that the two samples are duplicates.

Details

The returned probabilities are calculated from the likelihoods used throughout the rest of this package, by scaling them to sum to one across all possible relationships. For Complex='simp' these are PO=parent-offspring, FS=full siblings, HS=half siblings, GP=grand-parental, FA=full avuncular, HA=third degree relatives (incl half avuncular), and U=unrelated. For Complex='full' there are numerous double relationship considered (PO & HS, HS & HA, etc), making both numerator and denominator in the scaling step less unambiguous, and the returned probabilities an approximation.

The likelihoods are calculated by calling CalcPairLL once or twice for each id-dam and id-sire pair: once not conditioning on the co-parent, and once conditional on the co-parent, if any. For genotyped individuals this is done with focal='PO', and for dummy individuals with focal='GP'.

For relationships between a genotyped and a dummy individual, it may only be possible to determine that the genotyped individual is a second degree relative (GP, HS, or FA) to the dummy's offspring. This then results in a probability of at most 0.33, even when the two are indeed parent and offspring.

See CalcPairLL and the vignettes for further details.

Note that for large pedigrees this function can be fairly slow, especially when using CalcPairLL's default Module='ped' and Complex='full'.

Subsetting the genotype data may give different results, as the likelihoods and thus the probabilities depend on the allele frequencies in the sample.

Value

the Pedigree dataframe with the three applicable columns renamed to id-dam-sire, and 7 additional columns:

Probdam	Probability that individual in dam column is the maternal parent, rather than otherwise related (LL(PO)/sum(LL))
Probsire	Analogous for sire
Probpair	Probability for id-dam-sire trio. Approximated as the minimum of dam conditional on sire and sire conditional on dam, thus not including e.g. both being siblings (those other configurations are considered by sequoia during pedigree reconstruction, but can (currently) not be accessed directly)
dam_alt, sire_alt	Most likely alternative (not PO) relationship between id-dam and id-sire, respectively
Probdam_alt, Probsire_alt	Probability of most likely alternative relationship

See Also

CalcPairLL, LLtoProb

Examples

test_ped <- Ped_griffin[21:25,]
# add an incorrect sire to illustrate
test_ped$sire <- as.character(test_ped$sire)
test_ped$sire[5] <- 'i057_2003_M'
Ped_with_probs <- CalcParentProbs(test_ped, Geno_griffin)
print(Ped_with_probs, digits=2)
# Any non-genotyped non-'dummifiable' individuals are automatically skipped

# To get likelihoods for 'all' relationships, not just probabilities for
# PO & (next-)most-likely:
LL_sire_single <- CalcPairLL(
  Pairs = data.frame(id1=test_ped$id,
                     id2=test_ped$sire,
                     dropPar1='both', # drop both -> id2 as single parent
                     focal='PO'),
  Pedigree = Ped_griffin,   # pedigree to condition on
  GenoM = Geno_griffin, Plot=FALSE)

sequoia documentation built on Aug. 21, 2025, 5:43 p.m.