Are COI sequences the most appropriate data?
It is certainly imperfect to use COI sequences for population estimates of IGD. We gave several reasons in our paper (page 65), including the inability of mtDNA to reflect neutral evolutionary patterns, and the potential sensitivity of patterns to the minimum number of sequences per population (end of page 57).
We are aware of the inconsistent archiving practices among data-depositors (i.e., uploading only unique sequences to repositories) and we were careful not to overextend our conclusions because of this. Paz-Vinas et al.’s recommendation to use haplotype diversity/richness rather than pairwise nucleotide diversity is valid given the unknown sampling effort behind archived sequences.
As a community, we must decide whether these limitations preclude the use of COI in macrogenetics. No other marker offers such broad taxonomic and spatial scope. Studies using mtDNA reveal large-scale patterns in IGD with respect to latitude and human impacts (e.g., Miraldo et al. 2016); refining analyses of these existing macrogenetic datasets to confirm the existence of patterns is necessary to advance the field. There is an opportunity to gain further insight into IGD patterns by complementing large-scale analyses with fine-scale population genetic estimates for data-rich taxa.