cellxgene-census-annotation

You can transfer cell-type annotations to scRNA-seq data without downloading massive reference datasets by using cloud-based access to the CELLxGENE Census. This Skill queries >33M annotated cells directly via pre-computed scVI and Geneformer embeddings.

Yes, you can annotate query scRNA-seq cells using nearest-neighbor majority-vote against specific tissue references. The workflow supports tissue-specific reference lookup and filtering via TileDB value filters to map accurate predictions to your query metadata.

Yes, pre-computed scVI or Geneformer embeddings are essential for the cross-atlas vector search and label-transfer workflow. These embeddings enable cloud-based Census access and facilitate rapid querying without downloading large datasets.

The best way to filter duplicates and map cell-type predictions to query metadata is by following the end-to-end reproducible workflow provided. It guides you through querying the Census, filtering duplicates, and accurately mapping annotations to your query data.

Yes, CELLxGENE Census annotation works seamlessly with cloud-based TileDB value filters for querying single-cell data. This allows you to perform tissue-specific reference lookup and nearest-neighbor majority-vote annotation directly without local downloads.