Limitations¶

RustScenic is alpha research software. Use it deliberately and check outputs.

Current Limitations¶

• Adoption is early. The project has external-user reports, but not broad community use.
• The GRN implementation is not bit-identical to arboreto. Fine-grained edge rankings can differ.
• Topic modelling has algorithmic tradeoffs. Collapsed Gibbs improves topic diversity, while Mallet remains a strong reference.
• Full SCENIC+ parity on region-ranking databases still needs more real-data comparison.
• Some validation claims are currently based on canonical TF recovery, not full biological replication.
• A formal methods paper or preprint does not exist yet.

What Not To Claim¶

Do not claim:

• "RustScenic is universally better than pySCENIC."
• "RustScenic fully replaces every SCENIC+ analysis without caveats."
• "GRN edge rankings are identical to arboreto."
• "Community validation is the same as independent publication."

The accurate claim is narrower and stronger:

RustScenic is an installable, deterministic, CPU-focused implementation of the practical SCENIC compute path, with measured agreement on key downstream outputs and real-data validation evidence.