Running several NGS HLA typers and cross-checking their calls is best practice, but every typer speaks a different dialect - different file formats, different resolutions (2-field vs 3/4-field vs G-group), and the subtle one: they're often built against different IPD-IMGT/HLA releases, where allele names change over time. In practice that means renaming alleles by hand and building concordance tables in a spreadsheet.

hlaconcord does it reproducibly: parses each typer's native output > normalizes to canonical IPD-IMGT/HLA nomenclature > validates against a pinned release > reconciles database-version skew by matching renamed alleles on their stable accession id, so a version difference never shows up as a false discordance. Output is a per-locus concordance/consensus report (+ a GL String) with the genuine disagreements flagged.

Supported today: OptiType, arcasHLA, HLA-LA, HLA-HD (adding a typer is an isolated parser). No runtime dependencies - the in-house nomenclature reducer is stdlib-only and is cross-checked against py-ard as a test-time oracle in CI.

pip install hlaconcord

Code + docs: https://github.com/leonbzt/hlaconcord

It's early (v0.1.0) and not clinically validated - a decision aid, not a diagnostic. Feedback and typer requests very welcome.