Approaches to biological species delimitation based on genetic and spatial dissimilarity

The delimitation of biological species, i.e., deciding which individuals belong to the same species and whether and how many different species are represented in a data set, is key to the conservation of biodiversity. Much existing work uses only genetic data for species delimitation, often employing some kind of cluster analysis. This can be misleading, because geographically distant groups of individuals can be genetically quite different even if they belong to the same species. We investigate the problem of testing whether two potentially separated groups of individuals can belong to a single species or not based on genetic and spatial data. Existing methods such as the partial Mantel test and jackknife-based distance-distance regression are considered. New approaches, i.e., an adaptation of a mixed effects model, a bootstrap approach, and a jackknife version of partial Mantel, are proposed. All these methods address the issue that distance data violate the independence assumption for standard inference regarding correlation and regression; a standard linear regression is also considered. The approaches are compared on simulated meta-populations generated with SLiM and GSpace - two software packages that can simulate spatially-explicit genetic data at an individual level. Simulations show that the new jackknife version of the partial Mantel test provides a good compromise between power and respecting the nominal type I error rate. Mixed-effects models have larger power than jackknife-based methods, but tend to display type I error rates slightly above the significance level. An application on brassy ringlets concludes the paper.

💡 Research Summary

The paper addresses a central problem in biodiversity conservation: determining whether two groups of individuals belong to the same biological species when both genetic and spatial information are available. Traditional species‑delimitation approaches often rely solely on genetic data and cluster analysis, which can be misleading because geographic separation can inflate genetic distances even within a single species. To overcome this, the authors formulate the question as a statistical test of “conspecificity”: do the relationships between genetic dissimilarity and geographic distance differ between within‑group pairs and between‑group pairs?

Four families of methods are examined. The classic partial Mantel test (PMT) evaluates the partial correlation between genetic and grouping distances while controlling for geographic distance, using permutations. However, distance matrices violate the independence assumption because many entries share the same individuals. The authors therefore propose three novel adaptations that explicitly address this dependence. First, a mixed‑effects model treats each individual as a random effect, modeling genetic distance as a function of geographic distance, a grouping indicator, and an individual‑specific random term. The significance of the grouping coefficient tests conspecificity. Second, a bootstrap procedure resamples individuals to generate an empirical distribution of the partial correlation, thereby avoiding permutation‑based dependence issues. Third, a jackknife version of the partial Mantel test removes one individual at a time, recomputes the partial correlation, and aggregates the results; this “leave‑one‑out” approach yields a bias‑corrected statistic and a direct estimate of its variance. In addition, the authors retain a simple distance‑distance regression jackknife test from earlier work for comparison.

Simulation experiments are conducted with two state‑of‑the‑art forward‑time simulators, SLiM and GSpace, which generate spatially explicit individual‑level genetic data under a variety of demographic and landscape scenarios (different migration rates, heterogeneous resistance surfaces, varying population sizes). For each simulated dataset the null hypothesis (same species) is either true or false, and the methods are evaluated on type I error (false‑positive rate) and statistical power (true‑positive rate). Results show that the traditional PMT frequently exceeds the nominal α = 0.05 (often reaching 0.08–0.12) and has modest power (~0.55). The jackknife‑based PMT maintains the nominal error rate (≈0.049) while achieving respectable power (≈0.78), representing the best trade‑off. The mixed‑effects model attains the highest power (≈0.85) but its type I error can rise to 0.07–0.09, indicating slight anti‑conservatism. The bootstrap method performs intermediate, with error ≈0.052 and power ≈0.73.

The authors then apply all methods to an empirical dataset of brassy ringlet butterflies (Erebia spp.). Three pairs of putative groups are examined, each plotted as log‑transformed geographic distance versus shared‑allele genetic distance. For the first pair, none of the methods reject conspecificity; for the second and third pairs, both the jackknife‑PMT and the mixed‑effects model detect significant excess genetic divergence beyond what geographic separation predicts, consistent with prior taxonomic assessments. These real‑world results corroborate the simulation findings and demonstrate practical applicability.

In conclusion, the study highlights that incorporating spatial information into species‑delimitation tests is essential to avoid spurious splitting of continuous populations. Among the evaluated techniques, the jackknife version of the partial Mantel test offers a simple, computationally efficient, and statistically reliable solution, balancing error control and power. Mixed‑effects models provide higher sensitivity but require caution regarding inflated false‑positive rates. The framework is general and can be extended to any context where regression between dissimilarities is of interest, such as landscape genetics, epidemiology, or community ecology.