One API, AnnData throughout
Bulk, single-cell, spatial and multi-modal data share the same calls and conventions.
About
A unified Python ecosystem for multi-omics, with Rust and Torch underneath.
One library exposes the API. Around it, the organization rebuilds trusted R methods in Python and Rust, verifies them against the originals, and makes everything runnable by analysis agents.
Rust and Torch underneath
Hot loops in Rust, tensor methods on Torch, classical statistics on NumPy and SciPy.
Verified, not approximated
Ports are gated on numerical parity with the R reference before release.
Agent-ready
omicverse-skills and omicclaw let agents run the stack.
Governance
Track maintainers
Ports, runtime, agents and docs each have a named maintainer who reviews and releases.
Open review
Design discussion happens in GitHub Issues and Discussions. No maintainer status needed to contribute.
RFC for breaking changes
Public API, AnnData contract or benchmark changes go through a public RFC.
Code of conduct
Contributor Covenant. Reports go to the maintainer contact in the org README.
Release cadence
Stable monthly, patches as needed, each with changelog and parity diffs.
Independent
Funding and research-group acknowledgments live next to the code they support.
Citation
If you use OmicVerse in published work, please cite the framework paper. For a specific port, also cite the original method.
@article{omicverse, title = "OmicVerse: a unified multi-omics analysis framework", author = "Zeng, Zehua and others", journal = "Nature Communications", year = "2024", url = "https://github.com/omicverse/omicverse" }
Replace this entry with the canonical citation from the project README before publishing.