Back

Species delimitation using genome-wide estimates of Dxy and Fst

de Jong, M. J.; Wambach, C.; Muloongo, M.; Ferrette, B.; Wolf, M.; Janke, A.

2025-12-11 genomics
10.64898/2025.12.08.692692 bioRxiv
Show abstract

Accurate and stable species delimitation is essential for effective nature conservation and management. Advances in genomics now offer the potential for consistent decision-making. We performed a meta-analysis of genomic data from hundreds of sister lineages of large mammals to test whether species boundaries can be identified using genome-wide sequence dissimilarity estimates. We find that a combined threshold of absolute (Dxy = 0.225% [0.20%-0.27%]) and relative genetic distance (Fst = 0.26 [0.19-0.54]) predicts current taxonomic status with over 90% accuracy. This predictive power indicates that mammalian taxonomists managed to maintain consistency across taxa despite having to rely on disparate morphological traits. Lineage pairs exhibiting Haldanes Rule exceed far higher thresholds (Dxy = 0.4%; Fst = 0.55), implying that mammalian taxonomists adhere to species concepts which allow for interbreeding. Our findings challenge the significance of fast-track bottleneck speciation and suggest instead that deep divergences of up to 100,000 generations are usually required for mammalian sister lineages to remain genetically isolated upon secondary contact. We discuss taxonomic revisions to improve temporal banding, including three potential cases of cryptic speciation: southern versus eastern aardwolves, Asiatic black bears versus Japanese black bears, and European versus Asian wild boar. In summary, our meta-analysis provides simple rules for species delimitation and offers new insights into the taxonomy and speciation dynamics of large mammals.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.