Modernize to pip-installable package (v2.0)

[jonasscheid]

Summary

Transforms OptiType into a modern, pip-installable Python package.

Key changes:

• Package structure: src-layout with pyproject.toml (hatchling)
• CLI: Click-based with subcommands (optitype run, check-deps, init-config)
• Data format: HDF5 → Parquet (92MB → 370KB)
• Python: 3.10+ only (removed Python 2.7 compatibility)
• CI/CD: GitHub Actions with PyPI trusted publishing
• API: Programmatic interface (run_hla_typing())

Installation:

pip install optitype

Usage:

optitype run -i reads_1.fq -i reads_2.fq --dna -o results/

Results verified against baseline - HLA typing output unchanged.

Exome (1.3.5 top, 2.0.0 botton)

RNA (1.3.5 top,, 2.0.0 botton)

🤖 Generated with Claude Code

[jonasscheid]

@andras86 @b-schubert @lkuchenb @christopher-mohr

Would be great to have a modern optitype package. I would like to ask you to give feedback and help pushing a new optitype version to pypi. On top of this, new small features can be simply added, e.g. reference update etc.

[jonasscheid]

@andras86 @b-schubert @lkuchenb @christopher-mohr do not want to be annoying, but would be great to have your opinion on this PR and how to proceed. Lots of people are still using Optitype, would be a pity to discontinue maintenance. I can offer to help out with this as well.

[Zaroganos]

Following; mean to test this revision soon. Will report back. If this works directly in a python env this sounds promising.

[jonasscheid]

@Zaroganos sounds good! You can use the feature branch with bam/sam input in python env. With fastq input you require razer3, which is available via conda, so you might wanna switch to a conda env and run optitype from within there

[andras86]

Hi @jonasscheid!

Thank you for this, it's nice to see you want to keep OptiType alive, and I definitely won't prevent you from doing that :) You're right that in its original form it's a total PITA to use, so modernizing it to play along with the ecosystem would do great service to any users.

I'm okay with trusting you and whoever else is interested to test it, so if @Zaroganos is satisfied we can merge. The one thing I'd ask you though is to version it as v1.5 instead of 2.0, which was meant to be the version that types HLA Class II as well. A preliminary form of it (different concept from OT1) in some branch was used in a couple of projects (mostly ancient DNA), but let sleeping dogs lie.