giftag

giftag annotates genome and protein FASTA files with the markers gifter evaluates. It writes the gene-by-marker table gifter reads, together with search evidence and a record of which markers could be searched.

Read the giftag documentation for the quickstart, commands, output files, marker coverage, methods and validation, and the handoff to gifter.

Quickstart

Python 3.9 or newer is required. The full database build downloads about 2.9 GB from KOfam, NCBIfam, Pfam and dbCAN onto your machine and uses about 1.5 GB of disk space. Read the source terms before building.

pip install giftag
giftag build
giftag annotate -i genomes/ -o annotations/
giftag info

Then use the marker table in R:

markers <- read.delim("annotations/giftag_markers.tsv")
calls <- gifter::evaluate_gifts_community(markers)

Each FASTA file is one genome. giftag_markers.tsv contains genome_id, gene_id, namespace and accession for gifter, plus the profile, rule, score and threshold behind each accepted hit. The built database's markers.tsv identifies markers giftag could not search. Keep it and giftag_run.json with an analysis: an unsearched marker otherwise looks absent to gifter.

The tools make different claims. giftag observes marker evidence; gifter evaluates whether those observations complete a curated genomic capability. Neither tool infers expression, activity or phenotype from a hit.

Development

pip install -e '.[test]'
pytest

Documentation source is in docs/. From the repository root, run quarto render docs to build the site locally. The GitHub Pages workflow renders it on pull requests and deploys it from main.

The publication benchmark compares giftag with KofamScan, run_dbcan, direct HMMER and two InterProScan releases on a pinned eight-genome panel. Its reproduction guide includes the Mjolnir workflow, source hashes, environments and analysis scripts.

Licensing

giftag's code and original documentation are MIT-0. giftag redistributes no profile libraries. giftag build downloads them from their publishers, whose terms apply; see source terms.