Machine learning approaches delimit cryptic taxa in a previously intractable species complex.

Area of Science:

• Genomics
• Bioinformatics
• Evolutionary Biology

Background:

• Cryptic species lack morphological distinctions but are detectable via DNA.
• Existing genetic methods risk over-splitting species, especially in low-dispersal organisms.
• Machine learning (ML) offers a potential solution to improve species delimitation accuracy.

Purpose of the Study:

• To apply ML methodologies to delimit species within the Aoraki denticulata complex.
• To evaluate ML's effectiveness in handling extreme population structures in low-vagility species.
• To compare ML results with previous species delimitation studies on the same taxon.

Main Methods:

• Utilized a large dataset generated through hybrid target enrichment of ultraconserved elements (UCEs).
• Applied both unsupervised and supervised ML analyses for species delimitation.
• Incorporated training data from both broad animal groups and dispersal-limited analogues.

Main Results:

• ML approaches identified cryptic species with realistic geographic ranges, unlike previous studies.
• Results align with known distributions of morphologically diagnosable mite harvesters.
• ML proved effective in delimiting species in complexes with low-vagility cryptic species.

Conclusions:

• ML applied to UCE data is a robust method for species delimitation in low-vagility cryptic species.
• Using biologically relevant training data enhances the informativeness of ML analyses.
• This study validates ML as a superior tool for resolving complex species boundaries.