MetaWorks: A flexible, scalable bioinformatic pipeline for high-throughput multi-marker biodiversity assessments

biodiversity studies, different signature DNA regions are chosen for their ability to identify target taxa. For example, in prokaryotes, the 16S small subunit (SSU) ribosomal RNA (rRNA) region is often used for genus level taxonomic assignments [4,5]. Other popular markers include cyto- chrome c oxidase (COI) for animals; ribulose bisphosphate large subunit (rbcL) for plants and diatoms; the internal transcribed spacer (ITS) for fungi and plants; 18S SSU for eukaryotes, arbus- cular mycorrhizal fungi, and diatoms; and 12S mitochondrial SSU for fish [6–14]. Existing pipelines such as QIIME2 and DADA2 were initially developed to support the microbial ecology community [15,16]. In comprehensive, multi-trophic, multi-marker studies, there is a need for a pipeline that can handle rRNA genes, spacer regions, as well as protein- coding markers in a single harmonized environment [17,18]. For the ITS region, we needed a pipeline that could remove the conserved flanking rRNA genes as this has been shown to improve taxonomic assignment accuracy [19]. For protein-coding regions, we needed a pipe- line that could remove putative pseudogenes [20–23]. We also wanted the ability to generate high quality exact sequence variants (ESVs) for popular metabarcoding markers (not just 16S or ITS) for the additional level of genetic and taxonomic resolution ESVs can provide [24–26]. For taxonomic assignment, we wanted to use a classifier that would provide a measure of con- fidence for assignments to reduce false-positive assignments [27–29].

As multi-marker studies are carried out on phylogenetically divergent taxa, such as in bio- diversity or trophic studies, there is a need for more generic pipelines where different markers can be analyzed using similar dataflows with 3rd party programs instead of being limited to database-specific pipelines and tools [17,30]. We developed MetaWorks with the following objectives: 1) reproducibility with respect to the computational environment used as well as the pipeline itself, 2) scalability to leverage high performance computer clusters to speed up the analysis of large datasets, 3) naive Bayes classifier support for popular metabarcode mark- ers; and 4) to support marker-specific processing steps such as ITS extraction and pseudo- gene-removal for protein-coding markers. MetaWorks was designed for data analysts who are comfortable using Linux command-line tools but would like a single harmonized environment and pipeline to process multi-marker metabarcode datasets. Implementation and workflow Implementation MetaWorks is a multi-marker ‘meta’-barcode pipeline that does ‘the works’ by supporting the bioinformatic processing of popular markers including rRNA genes, spacers, and protein cod- ing genes generating taxonomically assigned ESVs or operational taxonomic units (OTUs). To facilitate reproducibility, scalability, and shareability of workflows we use the Conda package manager to facilitate the download of most programs and dependencies and the Snakemake workflow manager to automate pipelines and utilize computational resources efficiently [31–

33]. Snakemake supports re-entrancy and automatic deployment of multiple parallel jobs, both ideal for high performance computing environments where many cores are available to speed up the analysis of large datasets. We provided instructions on how to install and use Conda in the online documentation. One additional program not available as a Conda package, ORFfinder, may need to be down- loaded separately if pseudogene-filtering will be conducted and instructions are provided in the online documentation. MetaWorks can be downloaded from https://github.com/ terrimporter/MetaWorks and a suite of trained classifiers for taxonomic assignment are also available from GitHub (Table 1). Depending on the DNA metabarcode marker(s) the user will be processing, these can be individually downloaded from GitHub and instructions are pro- vided in the online documentation. PLOS ONE MetaWorks mult-marker bioinformatic pipeline PLOS ONE | https://doi.org/10.1371/journal.pone.0274260 September 29, 2022 2 / 11 terrimporter.github.io/MetaWorksSite. Additional trained classifiers that can be used by MetaWorks for taxonomic assignment are available from GitHub at https://github.com/terrimporter. Funding: MH received funding from Genome Canada and Ontario Genomics for the Sequencing the Rivers for Environmental Assessment and Monitoring (STREAM) project. TMP received funding from the Government of Canada through the Genomics Research and Development Initiative (GRDI), Metagenomics-based ecosystem biomonitoring (Ecobiomics) project. The funders had no role in study design, data collection and

analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.