Comment on Falade et al. (2016) DNA-barcoding of Clarias gariepinus, Coptedon zillii and Sarotherodon melanotheron from Southwestern Nigeria

DNA sequence data has become widely accepted as a useful tool for taxonomic determination and discovery^1–3. But the potential pitfalls of DNA taxonomy in operation have been forewarned for some time^4–10.

The DNA barcode itself is simply a standard region selected to facilitate comparison¹¹. A library built of many such sequences and based on a gene evolving at a rate that minimizes variation within and maximizes variation between species becomes a powerful taxonomic resource⁵. But the journey from DNA barcode sequence to species determination still requires critical application, particularly when applied to taxa or regions that are not currently well represented in sequence databases.

Falade et al.¹² obtained DNA sequences for sixteen individual fish from Southwest Nigeria, a region with relatively sparse coverage in sequence databases. Such data are valuable because broad geographic and taxonomic representation provide insight into genetic diversity within taxonomic groups and help us to refine hypotheses of species circumscription and phylogenetic relationships.

Falade et al.,¹² sequenced each specimen for the standard animal DNA barcode region cytochrome oxidase I (COI) and a region of the 16S mitochondrial ribosome. The authors queried their sequences against both the BOLD Systems (RRID: SCR_004278; boldsystems.org/index.php/IDS_OpenIdEngine) and NCBI GenBank (RRID: SCR_004860; BLASTN, RRID: SCR_001598; blast.ncbi.nlm.nih.gov/Blast.cgi) databases (because BOLD does not include 16S, these sequences were only compared to GenBank). Although the authors claim that “this resulted in straightforward identification”, we take a more nuanced view on their results.

The BOLD identification engine and BLASTN comparison with GenBank work differently and were created for different purposes^13–15; only BOLD is specifically intended to be used as a taxonomic identification tool, while BLASTN assesses sequence similarity. BLASTN will always return the most similar sequences in GenBank. BOLD is more discriminating, since it is limited to a handful of specific loci and uses similarity thresholds to assess whether or not a query sequence can be matched to identified sequences in the database with high confidence. BOLD will alert the user when it determines that no confident identification could be made. DNA-based identification is complicated by the fact that both BOLD and GenBank include misidentified sequences¹⁶.

BOLD failed to identify with confidence any of the sixteen COI sequences. Eight were classified as probably belonging to one of a handful of possible species, while the rest received no hit. From this, we infer that Falade et al. made their taxonomic determinations based almost entirely on BLASTN results. As reported (Table 1), all but one of these were scored as 98–99% identical to their top GenBank hit with the remaining sequence (KX231778; Coptodon_zilli_odooba_1) scoring 86% identical.

To view the results in context, we downloaded from BOLD all COI sequences identified as one of the three species specified by Falade et al. [search ‘Taxonomy’ for Clarias gariepinus, Sarotherodon melanotheron, and Coptodon zillii (the latter also under the synonym Tilapia zillii)]. These sequences were combined with the Falade et al. data and initially aligned using MAFFT version 7.187¹⁷ with manual adjustments made using Mesquite version 3.10¹⁸ (mesquiteproject.wikispaces.com/). A phylogenetic analysis was performed using RAxML version 8.2.8¹⁹. Initial alignment and phylogenetic analysis were performed through the CIPRES Science Gateway version 3.3²⁰ (RRID: SCR_008439; phylo.org/). Alignment required reversing or reverse-complementing some of the sequences from Falade et al. The problematic sequence KX231778 could not be satisfactorily aligned with the others and had to be excluded from the tree. The remaining COI sequences did cluster with other GenBank sequences in such a way as to suggest the remaining taxonomic determinations reported by Falade et al. are credible.

Another anomalous sequence is KX243287 (Clarias_gariepinus_asejire_12), a 16S sequence approximately twice the length of the others. We have no explanation for this.

The evidence presented by Falade et al. is not sufficient to determine at least the COI sequence KX231778. The method applied by Falade et al. made it nearly impossible to fail to obtain a taxonomic name for each sequence. This is a scientific flaw, and an example of the uncritical application of DNA taxonomy.

This paper was discussed as part of a regular journal discussion group offered by the Endless Forms research group at Naturalis Biodiversity Center, which involves students in the Evolution, Biodiversity, and Conservation program at Leiden University. Similar journal-article-based discussion groups can be found at many universities and Natural History Museums. We support the rationale behind open review journals (blog.f1000research.com/2014/05/21/what-is-open-peer-review/) and therefore decided to share the sense of our discussion with the broader community. We would like to encourage other journal discussion groups to include open review articles in their literature discussions, and consider sharing summaries of their discussions as article comments. Healthy science literature depends on a robust pool of potential reviewers²¹. We see journal discussion groups as an untapped resource for providing feedback on scientific literature, and also as incubators for developing student-scientists into constructive and rigorous peer reviewers.

Data availability

F1000Research: Dataset 1. Aligned COI sequence data, 10.5256/f1000research.9829.d141383²²

F1000Research: Dataset 2. Phylogenetic tree, 10.5256/f1000research.9829.d141384²³