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

Use case 2: Also as part of the STREAM project, MetaWorks results were used to analyze ESVs from diatoms (rbcL) and arthropods (COI) sampled within and across sites of varying water quality [54]. Using the MetaWorks pipeline, two different protein-coding markers were bioinformatically processed in two runs. The first run processed the rbcL marker, using a mix- ture of 5 different primers in a single adapters.fasta file, and pseudogenes were removed from this dataset using the simple ORFfinder method [21]. The second run processed three COI amplicons, each targeting an approximately 200 bp length of the COI barcoding region using 6 different primers in a single adapters.fasta file, and pseudogenes were removed from this data- set using the ORFfinder+HMMER method since a COI arthropod HMM model was available [21]. The study reported a diversity assessment across sites of varying water quality using rich- ness, effective richness, and beta diversity. Additionally, the taxonomic assignments generated from MetaWorks were used to obtain resource-consumer relationships from a global database of biotic interactions (GloBI) so that community stability using trophic and network measures could be assessed across sites with varying water quality [55]. This use-case shows how Meta- Works can handle a variety of protein-coding markers for trophic and network analyses to facilitate ecological assessments of freshwater condition. Use case 3: As a part of collaborative work with Environment and Climate Change Canada,

MetaWorks was used to assess macroinvertebrate and (non-macroinvertebrate) eukaryote taxa in an urban harbour using COI and 18S rRNA [56]. Using the MetaWorks pipeline, COI metabarcodes were identified down to species rank with 99% accuracy and 18S metabarcodes were identified to genus rank with 80% accuracy using a custom-trained classifier based on the SILVA 18S release 138 [38]. In this study, conventional macroinvertebrate sampling for assess- ing water quality in Toronto Harbour was compared with metabarcoding methods. COI meta- barcoding was found to detect more diversity at a finer level of taxonomic resolution compared with conventional approaches and was able to distinguish sites with particularly high levels of sediment contaminants. Additionally, the use of a multi-marker approach allowed microscopic eukaryote diversity to be sampled at the same time from the same sam- ples, producing indicators that responded to gradients in both sediment contaminants and water physical-chemical features. This use-case illustrates how MetaWorks can facilitate the application of multi-marker metabarcoding approaches that target different domains of life. As demonstrated in the above examples, MetaWorks supports a wide range of analysis sce- narios from metabarcoding data. We envision that MetaWorks will aid broader user commu- nities and fill a need in multi-marker metabarcoding studies that target taxa from multiple different domains of life, to provide a unified processing environment, pipeline, and taxo-

nomic assignment approach for each marker from ribosomal RNA genes, spacers, or protein coding genes. QIIME2 is perhaps the most popular and comprehensive platform for such work, but to date, focuses on processing mainly prokaryote and fungal datasets [16]. To our knowledge, MetaWorks is the only bioinformatic pipeline that can handle rRNA genes but that also integrates special processing steps to handle ITS spacers as well as filter out obvious pseudogenes in protein coding markers such as COI. There has been a lot of activity with respect to building new bioinformatic tools to handle COI metabarcodes. Recent work, such as the BOLDigger program, makes the BOLD identifi- cation engine more suitable for identifying large batches of COI metabaracodes and has both GUI and command-line interfaces for efficient sample processing [57]. A new program, called NUMTdumper, has been developed as a stand-alone program meant to be incorporated into bioinformatic pipelines [20]. NUMTdumper provides a method to screen for NuMTs based on read counts while acknowledging the trade-offs between removing all possible NuMTs while erroneously removing genuine reads. An R package called ‘coil’ has also recently been developed that will place COI barcode and metabarcode sequences in frame using profile PLOS ONE MetaWorks mult-marker bioinformatic pipeline PLOS ONE | https://doi.org/10.1371/journal.pone.0274260 September 29, 2022 7 / 11