Abstract
Phylogenomic subsampling is a method for studying the stability of phylogenetic analyses by taking random or ordered subsamples of loci and comparing phylogenetic analyses of those subsamples. The method is made possible by the large number of loci made available by application of next‐generation sequencing methods to phylogenetic questions. My laboratory has used phylogenomic subsampling to investigate the consistency and stability of various methods of phylogenetic analysis of multilocus data sets, including the so‐called species tree methods, which use the multispecies coalescent as a framework for interpreting gene tree heterogeneity. I was inspired to focus on this method for this symposium volume because my Harvard colleague Gonzalo Giribet had independently been using phylogenomic subsampling to explore various questions in invertebrate phylogenomics. Phylogenomic subsampling has many useful applications in phylogenomics, yet when reporting the particulars of the results of such analyses, care should be taken to focus primarily on discrepancies that achieve a high level of support by the bootstrap or other methods. Using a recently published example, I show that the methods used to summarize the results of a subsampling experiment, such as the threshold for reporting support for one or another tree or clade, can influence the perceived success or failure of concatenation or species tree methods. Single‐ versus double‐bootstrapping is also shown to produce different subsampling results. I suggest guidelines for analysing and reporting the results of phylogenomic subsampling and suggest that it should become a routine part of phylogenetic analysis in the next‐generation era.