Efficiency of model-based Bayesian methods for detecting hybrid individuals under different hybridization scenarios and with different numbers of loci

Accurate detection of offspring resulting from hybridization between individuals of distinct populations has a range of applications in conservation and population genetics. We assessed the hybrid identification efficiency of two methods (implemented in the <smallcapitals>structure</smallcapitals> and <smallcapitals>newhybrids</smallcapitals> programs) which are tailored to identifying hybrid individuals but use different approaches. Simulated first- and second-generation hybrids were used to assess the performance of these two methods in detecting recent hybridization under scenarios with different levels of genetic divergence and varying numbers of loci. Despite the different approaches of the methods, the hybrid detection efficiency was generally similar and neither of the two methods outperformed the other in all scenarios assessed. Interestingly, hybrid detection efficiency was only minimally affected by whether reference population allele frequency information was included or not. In terms of genotyping effort, efficient detection of F-1 hybrid individuals requires the use of 12 or 24 loci with pairwise F-ST between hybridizing parental populations of 0.21 or 0.12, respectively. While achievable, these locus numbers are nevertheless higher than the number of loci currently commonly applied in population genetic studies. The method of <smallcapitals>structure</smallcapitals> seemed to be less sensitive to the proportion of hybrids included in the sample, while <smallcapitals>newhybrids</smallcapitals> seemed to perform slightly better when individuals from both backcross and F-1 hybrid classes were present in the sample. However, separating backcrosses from purebred parental individuals requires a considerable genotyping effort (at least 48 loci), even when divergence between parental populations is high.