A phylogenetic approach to delimitate species in a probabilistic way

Abstract

Different species concepts and their associated criteria have been used to delimit species boundaries, such as the absence of gene flow for the biological species concept and the presence of morphological distinction for the morphological species concept. The need for different delimitation criteria largely reflects the fact that species are generated under various speciation mechanisms. A key question is how to make species delimitation consistent in a species group, especially when we want to delimit the species boundaries over many newly discovered evolutionary lineages and add these new lineages into a comparative analysis. Instead of forcing a single definition of ‘species’, we can acknowledge different delimitation criteria by modelling how fast lineages in a species group evolve to meet these criteria along a phylogenetic tree. This study presents such a new model and a new delimitation approach that calculates the probability of each possible species identity of a lineage. We use simulations to show that our likelihood function gives accurate estimates of parameters in the model and our approach have high power to correctly identify species identities. We apply the approach to lineages in two real species groups that already have genomic and morphological evidence for their species identities. Our approach gives consistent inference of species identities with these existing pieces of evidence. We also demonstrate how to use our model to test a popular hypothesis about speciation process across all lineages in a species group and discuss further extension of the model to study speciation.