Estimating Genetic Relatedness in Admixed Populations

Estimating genetic relatedness, and inbreeding coefficients is important to the fields of quantitative genetics, conservation, genome-wide association studies (GWAS), and population genetics. Traditional estimators of genetic relatedness assume an underlying model of population structure. Each individual is assigned to a population, depending on a priori assumptions about geographical location of sampling, proximity, or genetic similarity. But often, this population assignment is unknown and assumptions about assignment can lead to erroneous estimates of genetic relatedness. I develop a generalized method of estimating relatedness in admixed populations, to account for (1) multi-allelic genomic data, (2) including all nine Identity By Descent (IBD) states, and implement a maximum likelihood based estimator of pairwise genetic relatedness in structured populations, part of the software, InRelate. Replicated estimations of genetic relatedness between admixed full sib (FS), half sib (HS), first cousin (FC), parent-offspring (PO) and unrelated (UR) dyads in simulated and empirical data from the HGDP-CEPH panel show considerably low bias and error while using InRelate, compared to several previously developed methods. I also propose a bootstrap scheme, and a series of Wald Tests to assign relatedness categories to pairs of individuals.