UCE 4-layer

UCE, or Universal Cell Embeddings, is a zero-shot foundation model for single-cell gene expression data.

What it does

UCE embeds cells from AnnData datasets so they can be compared or visualized in a shared representation space. The official workflow focuses on .h5ad inputs.

Current status in Liatir

This model is visible as a preview. Install and run are not enabled until Liatir adds a dedicated runtime box for the UCE scripts, requirements, model files, and output contract.

Expected inputs

• AnnData .h5ad file.
• Species selection.
• Batch size and embedding options.

Expected outputs

• Cell embeddings.
• Intermediate processing files when needed.
• CSV/JSON summaries and provenance.

Hardware and installation

The official repository includes a PyTorch workflow. Model files can be downloaded by the official script, but Liatir still needs to manage those assets directly before exposing the model as installable.

Official source

• UCE on GitHub