Embpy Data: Static Gene and Protein Embeddings

Embpy Data contains static gene and protein embeddings packaged with the embpy library. The dataset includes 11 distinct embeddings covering a total of 201,717 indexed entities, as generated by Theislab on June 1, 2026. Embedding values are stored in Zarr arrays, with row identifiers, provenance, species metadata, and AnnData-like .uns metadata stored in sidecar files.

Use Cases

• Train machine learning models for gene function prediction based on static embeddings.
• Perform similarity searches across biological entities using the indexed embedding vectors.
• Integrate pre-computed embeddings into bioinformatics pipelines for downstream analysis.
• Benchmark new embedding methods against the provided static representations.

Strengths

• Includes provenance and species metadata for traceability.
• Contains 201,717 indexed entities across 11 embeddings.
• Follows a defined schema version (embpy.static_embedding_package.v1).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count per embedding is unknown, which may limit suitability assessment.

Provenance

Source

theislab

Collection Method

Packaged with the embpy library; specific generation method not detailed.

Freshness

Last updated 2026-06-01 15:31:44; freshness should be verified.