Scaling Up Species Delimitation From DNA Barcodes to Whole Organelle Genomes: Strong Evidence for Discordance Among Genes and Methods for the Red Alga Dasyclonium.

Molecular sequence data have become a ubiquitous tool for delimiting species and are particularly important in organisms where morphological traits are not informative about species boundaries. A range of statistical methods have been developed to derive species limits from molecular data, for example, by quantifying changes in branching patterns in phylogenetic trees. We aim to investigate how such methods scale up from single genes to whole organelle genomes. We gathered chloroplast genome data from 38 samples of the red algal genus Dascyclonium and analysed them with the popular species delimitation methods Assemble Species by Automatic Partitioning (ASAP), General Mixed Yule Coalescent (GMYC), and Poisson Tree Processes (PTP). We show extensive variation in inferred species boundaries depending on the method and dataset used. Genome-scale analyses differed substantially between methods, with ASAP predicting the fewest species, PTP intermediate, and GMYC inferring many species. Based on a series of simulations, we identify a tendency of GMYC to overestimate species numbers as alignments increase in length, while the other two methods are not sensitive to this scaling. Gene-by-gene analyses show strong differences in predicted species limits, which is unexpected seeing that all genes are on a single uniparentally inherited chromosome, and highlight that choosing a particular gene as a DNA barcode has significant consequences for species diversity estimates. We show extensive cryptic diversity in the genus Dasyclonium and propose a consensus solution for species limits based on our combined results, enriched with biogeographic and morphological interpretations. Finally, we make recommendations for interpreting the results and improving the inferences drawn from species delimitation methods.