Dr Jack Scanlan1, Dr Alexander Piper1,2, Prof Brendan Rodoni1,2, Dr Mark Blacket1,2
1Agriculture Science and Technology, Agriculture Victoria, Bundoora, Australia, 2School of Applied Systems Biology, La Trobe University, Bundoora, Australia
Biography:
Jack Scanlan is a Research Scientist at Agriculture Science and Technology (AST) within Agriculture Victoria, based at the AgriBio Centre in Bundoora, Victoria. He received his PhD from The University of Melbourne in 2020, where he studied the functional and evolutionary genomics of ecdysteroid kinase-like genes in insects. After a post-doc at The University of Melbourne studying the genomics of Leptosphaeria maculans, a fungal pathogen of canola, he moved to AST, where he works primarily on genomics and bioinformatics of agricultural pest insects. His current projects involve developing analysis and reference database curation pipelines for insect metabarcoding, population and functional genomics of grape phylloxera, and the use of RNA interference for agriculture pest control.
High-throughput sequencing of DNA barcodes is a powerful tool for ecological, biosurveillance, and biosecurity applications. However, it relies heavily on the availability of trustworthy and complete reference databases/libraries that accurately link sequences to taxonomic labels. Public barcode databases contain tens of millions of sequences, but their taxonomic annotations can be unreliable, incomplete, and impractical to manually verify. Here we present Mimir, a Nextflow-based bioinformatic pipeline designed to automate the routine curation of large, reliable DNA barcode databases, emphasising scalability, accessibility, and reproducibility. Mimir can download barcode-containing sequences (including mitogenomes) from GenBank and BOLD, incorporate trusted, user-generated sequences, and perform taxonomic harmonisation to maintain taxonomic consistency across input sources. Sequences can then be filtered based on length, amplicon region, nucleotide ambiguity, pseudogene characteristics, level of taxonomic classification, and redundancy. Mimir implements a novel and efficient taxonomic misclassification detection approach that first estimates minimum and maximum barcode identity thresholds for each taxonomic rank, then uses targeted exact clustering to find taxonomic outliers, while avoiding computationally intractable all-vs-all alignments. Mimir can also validate the presence and identifiability of key taxa in the final database, which is essential for biosecurity applications. We show that Mimir can curate more than 15 million public arthropod COI sequences into a compact, reliable database in less than 12 hours on an HPC system, and that a combined database performs better at taxonomic classification than GenBank- or BOLD-only databases. Future updates will focus on improving barcode flexibility beyond COI and formatting of databases for different taxonomic assignment tools.