eDNAFlow, an automated, reproducible and scalable workflow for analysis of environmental DNA sequences exploiting Nextflow and Singularity

Abstract

Metabarcoding of environmental DNA (eDNA) when coupled with high throughput sequencing is revolutionising the way biodiversity can be monitored across a wide range of applications. However, the large number of tools deployed in downstream bioinformatic analyses often places a challenge in configuration and maintenance of a workflow, and consequently limits the research reproducibility. Furthermore, scalability needs to be considered to handle the growing amount of data due to increase in sequence output and the scale of project. Here, we describe eDNAFlow, a fully automated workflow that employs a number of state-of-the-art applications to process eDNA data from raw sequences (single-end or paired-end) to generation of curated and noncurated zero-radius operational taxonomic units (ZOTUs) and their abundance tables. This pipeline is based on Nextflow and Singularity which enable a scalable, portable and reproducible workflow using software containers on a local computer, clouds and high-performance computing (HPC) clusters. Finally, we present an in-house Python script to assign taxonomy to ZOTUs based on user specified thresholds for assigning lowest common ancestor (LCA). We demonstrate the utility and efficiency of the pipeline using an example of a published coral diversity biomonitoring study. Our results were congruent with the aforementioned study. The scalability of the pipeline is also demonstrated through analysis of a large data set containing 154 samples. To our knowledge, this is the first automated bioinformatic pipeline for eDNA analysis using two powerful tools: Nextflow and Singularity. This pipeline addresses two major challenges in the analysis of eDNA data; scalability and reproducibility.

Benchmarking bioinformatic tools for fast and accurate eDNA metabarcoding species identification

“…The same comparison could be extended to recent bioinformatic programs, or programs not considered in this study, such as MeFit for merging and filtering (Parikh et al, 2016), or DUDE‐seq for denoising and correction of sequencing errors (Lee et al, 2017). We could also apply our comparison approach on other pipelines, such as eDNAFlow which produces ZOTUs and uses a LCA assignment method (Mousavi‐Derazmahalleh et al, 2021) or CoMa (Hupfauf et al, 2020).…”

The applicability of eDNA metabarcoding approaches for sessile benthic surveying in the Kimberley region, north‐western Australia

“…The application of the ITS2 assays to collected seawater samples in the Lalang-garram marine parks produced over 200 taxonomic assignments-the majority of which were assigned at a species level using an LCA algorithm (Mousavi-Derazmahalleh et al, 2021).…”

Secondary predation constrains DNA-based diet reconstruction in two threatened shark species

“…The LULU algorithm 44 was then run to curate the assignments assessing sequence similarity and their co-occurrence patterns with the default parameters: minimum_ratio_type = min, minimum_ratio = 1, minimum_match = 84, minimum_relative_cooccurence = 0.95. This entire process was completed on the Zeus SGI cluster based at the Pawsey Supercomputing Centre in Kensington, Western Australia using an abridged version of the fully automated ‘eDNAFlow' pipeline 45 . All raw sequencing data needed to replicate the study are available from Dryad Digital Repository: https://doi.org/10.5061/dryad.kd51c5b4s .…”