Comparing genome-based estimates of relatedness for use in pedigree-based conservation management

Researchers have long debated which estimator of relatedness best captures
the degree of relationship between two individuals. In the genomics era,
this debate continues, with relatedness estimates being sensitive to the
methods used to generate markers, marker quality, and levels of diversity
in sampled individuals. Here, we compare six commonly used genome-based
relatedness estimators (kinship genetic distance (KGD), Wang Maximum
Likelihood (TrioML), Queller and Goodnight (Rxy), Kinship INference for
Genome-wide association studies (KING-robust), and Pairwise Relatedness
(RAB), allele-sharing co-ancestry (AS)) across five species bred in
captivity–including three birds and two mammals–with varying degrees of
reliable pedigree data, using reduced-representation and whole genome
resequencing data. Genome-based relatedness estimates varied widely across
estimators, sequencing methods, and species, yet the most consistent
results for known first order relationships were found using Rxy, RAB, and
AS. However, AS was found to be less consistently correlated with known
pedigree relatedness than either Rxy or RAB. Our combined results indicate
there is not a single genome-based estimator that is ideal across
different species and data types. To determine the most appropriate
genome-based relatedness estimator for each new dataset, we recommend
assessing the relative: (1) correlation of candidate estimators with known
relationships in the pedigree and (2) precision of candidate estimators
with known first-order relationships. These recommendations are broadly
applicable to conservation breeding programs, particularly where
genome-based estimates of relatedness can complement and complete poorly
pedigreed populations. Given a growing interest in the application of wild
pedigrees, our results are also applicable to in-situ wildlife management.