On the independent loci assumption in phylogenomic studies

Studies using multi-locus coalescent methods to infer species trees or historical demographic parameters usually require the assumption that the gene tree for each locus (or SNP) is genealogically independent from the gene trees of other sampled loci. In practice, however, researchers have used two different criteria to delimit independent loci in phylogenomic studies. The first criterion, which directly addresses the condition of genealogical independence of sampled loci, considers the long-term effects of homologous recombination and effective population size on linkage between two loci. In contrast, the second criterion, which only considers the single-generation effects of recombination in the meioses of individuals, identifies sampled loci as being independent of each other if they undergo Mendelian independent assortment. Methods that use these criteria to estimate the number of independent loci per genome as well as intra-chromosomal \"distance thresholds\" that can be used to delimit independent loci in phylogenomic datasets are reviewed. To compare the efficacy of each criterion, they are applied to two species (an invertebrate and vertebrate) for which relevant genetic and genomic data are available. Although the independent assortment criterion is relatively easy to apply, the results of this study show that it is overly conservative and therefore its use would unfairly restrict the sizes of phylogenomic datasets. It is therefore recommended that researchers only refer to genealogically independent loci when discussing the independent loci assumption in phylogenomics and avoid using terms that may conflate this assumption with independent assortment. Moreover, whenever feasible, researchers should use methods for delimiting putatively independent loci that take into account both homologous recombination and effective population size (i.e., long-term effective recombination).