mimir is a Nextflow-based pipeline for reference database curation of DNA barcode sequences, such as those required for metabarcoding experiments. It is partially a successor to taxreturn.
This pipeline is being developed by a team at Agriculture Victoria Research, as a part of the National Grains Diagnostic & Surveillance Initiative (NGDSI).
mimir is a sibling pipeline to freyr, which handles sequencing data analysis for metabarcoding experiments.
This pipeline is currently UNFINISHED and being actively developed, with no guarantee that the code is stable or usable!
Best way to access Barcode of Life Data (BOLD) data at the moment is with a direct link to the latest complete database package. You'll need to create an account with BOLD, which will allow you to generate a URL that will download the complete database package (>2 GB compressed) for 24 hours. Input an active URL into the pipeline in the following format, making sure to wrap the URL in quotes: --bold_db_url "https://www.boldsystems.org/index.php/API_Datapackage?id=BOLD_Public.06-Sep-2024&uid=166f4b93030986"
If you have already downloaded a complete BOLD data package, either from a previous pipeline run or a manual download, you can supply a path to either the compressed .tar.gz file itself or a directory where the compressed or uncompressed file(s) are found, using the --bold_db_path flag.
TODO: Add ability for BOLD login details to be given to pipeline instead of generated URL.
We are also looking into ways of downloading sequence data from BOLD using the bold R package, but this method is inherently less reliable due to unpredictable rate-limiting from the BOLD website.
For the best compatibility between source databases (ie. Genbank and BOLD), only taxa from the following ranks should be used as --target_taxa: kingdom, phylum, class, order, family, genus, or species. This is because intermediate ranks, like "subfamily" and "tribe", are not available for all sequences. If a database comprising only an intermediate rank is needed, we recommend filtering a larger database using a known list of component taxa, or using such a list as the input to the pipeline.
Currently, the pipeline parses input taxa through the NCBI taxonomy, so --target_taxa must be a recognised NCBI taxon name (eg. "Insecta"), or even better, an NCBI taxid number (eg. "50557"). Taxids are preferred as they are unambiguous, while taxon names are sometimes shared between different taxa (a good example is "Drosophila" belonging to three taxids: 7215, 32281 and 2081351). To assist with (but not completely solve) taxon name disambiguation, --target_rank must also be used to supply the taxonomic rank of the taxon name (eg. "class" for "Insecta"); this also helps taxonomic harmonisation between the NCBI and BOLD taxonomies.
At the moment, only two markers are supported: COI and 28S. More markers will be available soon.
A key input into the pipeline is a profile hidden Markov model (PHMM) of your marker of interest, which is used to distinguish between 'on-target' and erroneous sequences in external (and internal) data sources. A model for insect COI is contained in this repository at ./assets/folmer_fullength_model.rds, but models for other markers need to be provided by the user. A method to generate a PHMM from high-quality internal sequences will be available soon.