Do marker-based paternity assignments favour heterozygous and unrelated males?

Genetic marker-based parentage analyses are widely applied to studies of natural populations in the fields of evolutionary biology, conservation biology and ecology. When the same markers used in a parentage analysis are used together with the inferred parentage in a downstream analysis, such as the analysis of mate choice in terms of heterozygosity or relatedness, a bias may be incurred because a subset of the genotypes are favoured in parentage assignments or non-exclusions. A previous simulation study shows that exclusion-based paternity analyses are biased in favour of heterozygous males, and males less related to the mothers than expected under random mating. In this study, I investigated the biases of genetic paternity analyses achieved by both exclusion- and likelihood-based methods, using both analytical and simulation approaches. It is concluded that while both exclusion- and likelihood-based methods can lead to biased paternity assignments or non-exclusions in favour of a subset of genotypes, the bias is not consistently towards heterozygous males or males apparently less related to mothers. Both the direction and extent of the bias depend heavily on the allele frequency distribution and the number of markers, the methods used for paternity assignments, and the estimators of relatedness. There exist important differences in the patterns of the biases between exclusion- and likelihood-based paternity analysis methods. It is concluded that the markers, except when they are highly informative to yield accurate paternity assignments or exclusions, should be split into two subsets which are used separately in the paternity and downstream analyses.