Why

Remote databases change. They add organisms, diseases and studies, they rename search fields, and occasionally they change the format of the files themselves. Code that hardcodes those details breaks quietly, producing empty result sets or mislabelled columns rather than errors.

immvoke keeps a checked-in snapshot of each remote source’s search schema and of its downloaded file format. The command line is generated from that snapshot, so no field list is hardcoded, and a scheduled job re-harvests the remote schema and opens a pull request whenever it drifts. Breakage shows up as a reviewable diff and a failing test, not as a silently wrong download.

Usage

Datasets (OAS), producing an airrflow samplesheet:

immvoke --version
immvoke oas download paired --species human --limit 1 --outdir tmp    # To convert to fasta, rerun (hashes any file already on disk)
immvoke oas download paired --species human --limit 3 --outdir tmp --format fasta
cd tmp
nextflow run nf-core/airrflow -r 5.1.0 \
    -profile docker \
    --mode assembled \
    --input samplesheet_airrflow_fasta.tsv \
    --outdir airrflow_out -c ../airrflow.config \
    --clonal_threshold 0.2 -resume

Germline references, producing the reference_base and IgBLAST databases:

# IMGT germline for a species, and (with --igblast) the IgBLAST databases
immvoke imgt download human --outdir ref --igblast

# the AIRR-C sets blended with IMGT (immunoglobulin from OGRDB, TR and the
# remaining constants from IMGT) -- the airrflow airrc-imgt reference
immvoke airrc-imgt download human --outdir ref --igblast

# validate a germline folder someone provided and build its databases;
# --check validates only, without makeblastdb
immvoke reference build ref/reference_base --out igblast_base --check

nf-core/airrflow uses the result either way: point --reference_fasta and --reference_igblast at ref/reference_base and igblast_base with --fetch_germlines none.

License

This work is licensed under the GNU Affero General Public License 3 (AGPL-3).