Digest: Species delimitation in the face of demographic processes*

In recent decades, molecular methods for species delimitation have been developed, often embedded within the multispecies coalescent (MSC) framework (Yang and Rannala 2010), which does not discriminate between incipient species and structured populations (Sukumaran and Knowles 2017). MSC approaches assume that gene flow stops when populations diverge and that shared genetic variation is due to incomplete sorting of ancestral polymorphisms, which is the case in allopatric speciation without gene flow. However, since gene flow seems to be common in nature, an approach that relies solely on divergence is not ideal (Mallet et al. 2016).

In this issue, Smith and Carstens (2019) introduce the R package delimitR, which identifies species boundaries in the presence of divergence, gene flow, and other demographic processes (Smith and Carstens 2019). delimitR provides researchers a flexible and efficient tool to investigate speciation events and species’ boundaries. This package leverages the researchers’ expertise of their system to create a set of models, including divergence times, gene flow, migration, population expansion, and species tree topologies, as parameters. delimitR uses the coalescent simulator fastsimcoal2.6 (Excoffier et al. 2013) to generate data using a coalescent framework under different divergence scenarios. This produces multidimensional site frequency spectra, which are then binned to coarsen their dimensionality. Next, the random-forest classifier, a machine-learning approach, uses these bins to assess how well each model's simulated data compares to empirical observations. This allows researchers to identify the divergence scenario that best explains the data. Notably, delimitR shows low rates of misclassification in simple simulations, and only slightly higher rates in systems with greater complexity (e.g. changes in population size and gene flow).

Smith and Carstens (2019) apply their new method to the reticulate taildropper slug (Prophysaon andersoni), which likely contains multiple undescribed lineages (Smith et al. 2018). Comparison of 208 speciation models supported a four-population model with secondary contact between two populations. The timing of this secondary contact, coupled with evidence of habitat divergence between cryptic taxa, implies that character displacement might have occurred after the Last Glacial Maximum (Pielou 2008). Results from BPP (Yang and Rannala 2010), a popular species delimitation method that does not take gene flow into account, also pointed to four distinct lineages within the reticulate taildropper slug. However, this approach did not highlight the importance of secondary contact in the study system. Analyses for other groups (Hemidactylus geckos, Exyra moths, and Peucetia spiders) confirmed previously established groups and supported gene flow at various time points in the latter two.

Previous divergence-only approaches to delimitation have been useful for systems in which speciation occurs allopatrically and without secondary contact; delimitR builds on them by taking into account population-level processes that might obscure recent speciation. Further, delimitR is intentionally agnostic to species concept in grouping, splitting, and describing processes occurring between taxa. Whether the groups and processes suggested by delimitR are in fact species is up to the user, which emphasizes that the package should be complemented by ecological and morphological data. The use of expert knowledge to infer species limits goes back to Charles Darwin's view on species in On the Origin of Species: “Hence, in determining whether a form should be ranked as a species or a variety, the opinion of naturalists having sound judgment and wide experience seems the only guide to follow.” However, in contrast to Darwin's time, the opinion of naturalists is not the only guide to follow. The continuous development and improvement of species delimitation methods will advance the inference of species limits.

Associate Editor: K. Moore

Handling Editor: T. Chapman