|
Abstract:
|
Estimating kinship from genetic data is a challenging endeavor even in homogeneous human populations. Starting from third degree relatives, the coefficients of variation for the kinship coefficients are greater than 1. The presence of additional genetic correlations further challenges many existing methods for relationship inference. Producing good estimates of recent genetic relatedness is important, however, for population and quantitative genetics, genome-wide association studies, among others. A complex high-resolution dataset derived from real publicly available genotypes is produced to have both distant and recent genetic relatedness. Specifically, admixture of three continental populations, discrete sub-populations, reflective of sub-continental variation, and also non-trivial, familial-type recent relatedness within each of the sub-populations are all present. Using this dataset with complex genetic correlations we challenge several methods for relationship inference that (1) rely on genotype data and (2) aim to be robust to the presence of population structure and/or admixture. We propose a novel approach to improve the estimates.
|
