Taxonomic identification accuracy from BOLD and GenBank databases using over a thousand insect barcodes from Colombia

Abstract

Recent declines of insect populations at high rates have resulted in the need to develop a quick method to determine their diversity and to process massive data for the identification of species of highly diverse groups. A short sequence of DNA from COI is widely used for insect identification by comparing it against sequences of known species. Repositories of sequences are available online with tools that facilitate matching of the sequences of interest to a known individual. However, the performance of these tools can differ. Here we aim to assess the accuracy in identification of insect taxonomic categories from two repositories, BOLD Systems, and GenBank. This was done by comparing the sequence matches between taxonomist identification and the suggested identification from the platforms. We used 1160 sequences in eight orders of insects from Colombia. After the comparison, we reanalyzed the results from a representative subset of the data from the subfamily Scarabaeinae (Coleoptera). Overall, BOLD systems outperformed GenBank, and the performance of both engines differed by orders and other taxonomic categories (species, genus and family). Higher rates of accurate identification were obtained at family and genus level. The accuracy was higher in BOLD for the orders Coleoptera at family, for Coleoptera and Lepidoptera at genus, and species level. The other orders performed similarly in both repositories. Moreover, the Scarabaeinae subset showed that in this group species were correctly identified when BOLD match percentage was above 93.4% and a total of 85% of the samples were correctly assigned to a taxonomic category. These results accentuate the great potential of the identification engines to place insects accurately into their respective taxonomic categories based on DNA barcodes and highlight the reliable use of BOLD Systems for insect identification in the absence of a large reference database for a highly diverse country.

DNA barcoding insufficiently identifies European wild bees (Hymenoptera, Anthophila) due to undefined species diversity, genus‐specific barcoding gaps and database errors

“…We then transformed this frequency distribution into relative values to get a straightforward observation of the intra‐ and inter‐specific overlap proportions. To test for a “true” barcoding gap, that is, not obscured by obvious misidentifications, it is sometimes suggested to trim the distance dataset, thereby removing the effect of erroneous sequences that are abundant in public repositories and often represent well over 5% of all sequences (Baena‐Bejarano et al., 2023; Čandek & Kuntner, 2015; Meier et al., 2006, 2008; Meiklejohn et al., 2019; Wu et al., 2021). We trimmed roughly equal number of distances at both extremes (largest intraspecific and smallest interspecific) to open a 2% wide barcoding gap (Figure 2).…”