Evidence-based strategies to navigate complexity in dietary DNA metabarcoding: A reply

Abstract

It is clearly beneficial to eliminate low-abundance sequences that arise in error during dietary DNA metabarcoding studies, but to purge all low-abundance sequences is to risk eliminating real sequences and complicating ecological analyses. Our prior literature review noted that DNA sequence relative read abundance (RRA) thresholds can help ameliorate false-positive taxon occurrences, but that historical emphasis on this utility has fostered uncertainty about the associated risk of inflating the false-negative rate (Littleford-Colquhoun et al., 2022). To address this, we combined a simulation study and an empirical data set to both illustrate the issue and provide blueprints for simulation studies and sensitivity analyses that can help investigators avoid overcorrecting and thereby bolster confidence in ecological inferences. Awareness of both the costs and the benefits of abundance-filtering is needed because accurately characterizing dietary distributions can be critically important for understanding animal diets, nutrition and trophic networks. Highlighting the need to raise awareness, a critique of our paper emphasized the misleading notion that "false positive interactions between species can present fundamentally incorrect network structures in network ecology, whereas false negatives will provide a correct but incomplete version of the network" (Tercel & Cuff, 2022). Asserting that the reliability of results will be eroded by false positives but resilient to the omission of true positives is risky and runs counter to evidence. Unfortunately, abundance-filtering methods can introduce false negatives at higher rates than they eliminate false positives and thereby undermine the analysis of otherwise reliable sequencing data. Overcorrecting can qualitatively alter and ultimately undermine ecological interpretations.