Algorithmic single-locus species delimitation: effects of sampling effort, variation and non-monophyly in four methods and 1870 species of beetles

The vast number of undescribed species and the fast rate of biodiversity loss call for new approaches to speed up alpha taxonomy. A plethora of methods for delimiting species or Operational Taxonomic Units (OTUs) based on sequence data have been published in recent years. We test the ability of four delimitation methods (BIN, ABGD, GMYC, PTP) to reproduce established species boundaries on a carefully curated DNA barcode dataset of 1870 North European beetle species. We also explore how sampling effort, variation within species, nearest neighbor divergence and non-monophyly affect the OTU delimitations. All methods produced approximately 90% identity between species and OTUs. Effects of variation and sampling differed between methods. ABGD was found to be sensitive to singleton sequences, while GMYC showed tendencies for oversplitting. The best fit between species and OTUs was achieved by using simple rules to find consensus between discordant OTU delimitations. Using several different approaches simultaneously allows the methods to compensate for each other’s weaknesses. Barcode-based OTU-picking is an efficient way to delimit putative species from large datasets where the use of more sophisticated methods based on multi-locus or genomic data is not feasible.