The effects of sampling on delimiting species from multi-locus sequence data

Area of Science:

• Evolutionary Biology
• Genetics
• Bioinformatics

Background:

• Species delimitation is fundamental to life sciences but challenging due to cryptic species and incomplete lineage sorting.
• Accurate species delimitation requires robust methods using multi-locus DNA sequence data.

Purpose of the Study:

• To evaluate the accuracy and sampling sensitivity of three novel species delimitation methods: Structurama, Gaussian clustering, and nonparametric delimitation.
• To compare these methods using simulated multi-locus DNA data across varying species divergence levels and sampling strategies.

Main Methods:

• Simulated multi-locus DNA sequence data with two species tree depths.
• Tested sampling strategies from 5 alleles/5 loci to 20 alleles/100 loci.
• Evaluated Structurama, Gaussian clustering, and nonparametric delimitation without a priori sample assignments.

Main Results:

• Structurama accurately delimited recently diverged species ( > 1.5 N generations) with ≥10 loci.
• Gaussian clustering performed well for deeply divergent species ( > 2.5 N generations) but lacked sensitivity for recent divergences.
• Nonparametric delimitation required ≥25 loci and accurate gene trees, often over-splitting species with estimated genealogies.

Conclusions:

• Structurama is a powerful tool for species delimitation from multi-locus DNA data.
• Gaussian clustering and nonparametric methods have limitations in accuracy and sensitivity.
• Integrating multiple methods and data types is recommended for comprehensive species delimitation.