omicverse-cross-modal-celltype-transfer

Transfer cell-type labels from a reference AnnData to a query AnnData using weighted KNN mapping on a shared embedding basis. The Skill writes predicted labels and uncertainty scores directly back into the query object for downstream analysis.

Yes, weighted KNN supports cross-modal annotation workflows on shared latent embeddings such as X_glue. You can label ATAC cells from an RNA reference or reuse the workflow for other shared-embedding transfer tasks within OmicVerse pipelines.

You need a reference AnnData and a query AnnData that share a common embedding basis, along with prefix-matched label columns. The Skill requires this shared latent space to execute the weighted KNN mapping and calculate uncertainty.

Yes, the cell type transfer workflow outputs uncertainty scores alongside predicted labels. These metrics are written directly into the query AnnData object, allowing you to assess transfer confidence during downstream analysis.

Visualize transferred cell type labels using optional plotting and MDE-based visualization. This requires the necessary embedding basis to be available in the query AnnData, and the pymde package must be installed for MDE visualization.

Weighted KNN cross-modal transfer depends entirely on a precomputed shared embedding basis and prefix-matched label columns. Without a high-quality shared latent space between the reference and query AnnData, the label transfer and uncertainty calculations will not be reliable.