Jared A Grummer, Eric C Anderson, Gordon Luikart
Hybridization is ubiquitous throughout the tree of life. Its presence in genetic datasets often violates model assumptions or leads to erroneous inferences. Nonetheless, few parentage assignment methods have been developed that account for hybridization or admixture, particularly in non-model systems. Here, we describe a novel method of parentage analysis, MixedUpParents, that accounts for hybridization by using species-diagnostic single nucleotide polymorphisms while accounting for mixed ancestry within discrete genomic segments. We test MixedUpParents alongside other parentage assignment programs through simulations to assess assignment accuracy in scenarios with and without hybridization. Our simulations included a range of parameter values, including missing genotypes, migration (hybridization) rates and proportions of un-sampled individuals. Simulation results showed that in all scenarios with admixture, MixedUpParents outperformed the other parentage methods tested and performed well for parent-offspring assignment even when only one parent is sampled. In the best cases, MixedUpParents achieved ≥ 95% true-positive and ∼ 5% false-positive rates (i.e., assigning the wrong parent). Accuracy of all parentage methods increased when a higher proportion of the population was sampled, but decreased with higher levels of missing marker data. The most challenging cases for all methods were those with intermediate migration rates, and when full-siblings were included in the parent sample (representing age-agnostic field sampling). Overall, we show that 'traditional' methods can yield high false-positive rates when hybridization is present, emphasizing the importance of using appropriate methods in admixed systems. Our results and freely available software should advance population management and research on measures of reproductive success (fitness) in wild and captive populations, and parentage and pedigree reconstruction broadly.