Model-Based Species Delimitation: Are Coalescent Species Reproductively Isolated?

Abstract

A large and growing fraction of systematists define species as independently evolving lineages that may be recognized by analyzing the population genetic history of alleles sampled from individuals belonging to those species. This has motivated the development of increasingly sophisticated statistical models rooted in the multispecies coalescent process. Specifically, these models allow for simultaneous estimation of the number of species present in a sample of individuals and the phylogenetic history of those species using only DNA sequence data from independent loci. These methods hold extraordinary promise for increasing the efficiency of species discovery but require extensive validation to ensure that they are accurate and precise. Whether the species identified by these methods correspond to the species that would be recognized by alternative species recognition criteria (such as measurements of reproductive isolation) is currently an open question and a subject of vigorous debate. Here, we perform an empirical test of these methods by making use of a classic model system in the history of speciation research, flies of the genus Drosophila. Specifically, we use the uniquely comprehensive data on reproductive isolation that is available for this system, along with DNA sequence data, to ask whether Drosophila species inferred under the multispecies coalescent model correspond to those recognized by many decades of speciation research. We found that coalescent based and reproductive isolation-based methods of inferring species boundaries are concordant for 77% of the species pairs. We explore and discuss potential explanations for these discrepancies. We also found that the amount of prezygotic isolation between two species is a strong predictor of the posterior probability of species boundaries based on DNA sequence data, regardless of whether the species pairs are sympatrically or allopatrically distributed. [BPP; Drosophila speciation; genetic distance; multispecies coalescent.]

Standardized nuclear markers advance metazoan taxonomy

“…Some of our study cases found an extraordinary congruence of USCO-based analyses with species hypotheses derived from previous integrative approaches with independent evidence (Supplementary Information). However, given the tendency of over-splitting, our results also underline the steady need for critical evaluation and additional refinement (when possible) of the results of species delimitation (16) with additional evidence and methodology (57,65,66), also when inferring species boundaries with data-rich multi-gene datasets under the multispecies coalescent model (18,19,60). Limitations of current species delimitation approaches (58), but also the nature of species and speciation (67), continue to urge for an integrative evaluation of results of species delimitation (14) and emphasize the need to further improve protocols, algorithms and implemented models for species delimitation based on DNA data.…”

Widespread polytypic species or complexes of local species? Revising bumblebees of the subgenus Melanobombus world-wide (Hymenoptera, Apidae, Bombus)

“…Under ideal conditions with sufficient sampling, species are expected to appear in trees as groups with short internal branches but with longer external branches between them. Coalescent species have been found to show a 77% correspondence to interbreeding species for Drosophila fruit flies, one of the groups studied most intensively in speciation studies (Campillo et al 2020).…”

Defining Relictual Biodiversity: Conservation Units in Speckled Dace (Cyprinidae:Rhinichthys osculus) of the Greater Death Valley Ecosystem

“…Discrete population-level patterns can indeed be identified, but an over-interpretation of these patterns is a management issue in that populations are often inappropriately categorized. These 84 fears are borne out as well through complex analytical techniques commonly used to delimit species (Campillo, Barley, & Thomson, 2019;Sukumaran & Knowles, 2017). In addition, 86 sophisticated analytical methods are now emerging, such as Machine Learning (ML), that employ pattern recognition as a mechanism to identify biodiversity, and which allow systems to 88 automatically learn and improve from experience without being explicitly so programmed.…”