Oral Presentation Society for Freshwater Science 2026 Annual Meeting

Separating the wheat from the chaff: How sensitive are biodiversity surveys and ecological assessments to the bioinformatics pipelines used in eDNA metabarcoding (136181)

Leonce Ngirinshuti 1 , Torrey Rodgers 1 , Charles Hawkins 1
  1. Utah State University, Logan, UTAH, United States

Metabarcoding of environmental DNA (eDNA) is increasingly used in both biodiversity surveys and ecological assessments. However, the metabarcoding process is susceptible to errors in assigning sequences to described species or identifying undescribed species (e.g., operational taxonomic units – OTUs). These errors are often caused by either (1) field and laboratory contamination, (2) errors during PCR amplification due to primer biases, or (3) formation of artifactual sequences during sequencing, which can result in both false positive and negative detections. Bioinformatics pipelines differ in how they remove sequence errors and the tradeoffs associated with retaining erroneous sequences and removing true biological sequences. To date, only a few studies have examined these tradeoffs and their subsequent effects on both biodiversity surveys and ecological assessments. In this study, we addressed two specific questions: (1) to what extent do the different noise filtering methods used in two different pipelines affect estimates of taxa richness and composition measured as operational taxonomic units (OTUs)? (2) does choice of pipeline influence the strength of OTU-environment relationships? To test these ideas, we used two bioinformatic pipelines to identify the OTUs present in 751 eDNA samples collected from streams in eastern Washington, USA: the Ribosomal Database Project (RDP) classifier and the High-Accuracy Pipeline for Processing deep metabarcoding data (HAPP). The RDP relies on the DADA2-denoising process, and the HAPP uses both DADA2 and additional filtering steps designed to remove low quality sequences and chimeras. The raw OTU output from both pipelines contained substantial errors which greatly inflated richness. We then removed additional likely errors by dropping singleton OTUs. The RDP identified more OTUs (345 OTUs) than HAPP (257 OTUs). However, RDP- and HAPP-based NMDS ordinations were similar to one another, and the direction and strength of longitude, latitude, and temperature vectors overlaid on OTU ordinations were nearly identical (R2 = 0.61, 0.74, and 0.01 for RDP and 0.6, 0.68, and 0.01 for HAPP). The metabarcoding community needs standardized methods for quantifying erroneous OTUs, and practitioners need to be both transparent in reporting metabarcoding errors and aware of how these errors can influence estimates of biodiversity and assessments of ecological condition.