Sampling strategies for species trees: the effects on phylogenetic inference of the number of genes, number of individuals, and whether loci are mitochondrial, sex-linked, or autosomal.

Systematists can now use multi-locus data to construct species trees that take into account the stochastic nature of gene tree divergence among populations. There is a need to evaluate the new methods for species tree reconstruction in order to determine what kinds of loci to use and the most effective sampling schemes in terms of numbers of genes and numbers of individuals per species. Here we study sampling strategies with an empirical data set for six shorebird species in which we sequenced 1 mitochondrial, 12 autosomal, and 12 Z-linked loci for >8 individuals/species. We found that sampling greater numbers of genes resulted in substantial improvements to the resolution of the species tree, but sampling greater numbers of individuals had minor effects. We found that Z-linked loci significantly outperformed autosomal loci at all levels of sampling, which likely resulted from the lower effective population size of the Z-linked loci. Therefore, sex-linked loci are likely to be a powerful tool for multi-locus phylogenetic studies. We found that adding a mitochondrial gene to a set of Z-linked or autosomal loci substantially improved the resolution of the tree. Overall, our results help evaluate how best to maximize phylogenetic resolution while minimizing the costs of sequencing and computation when performing species tree analyses.