cellxgene-census

To query single-cell data across millions of cells, you programmatically access the CELLxGENE Census to retrieve standardized metadata and expression matrices from thousands of datasets. It supports filtering by is_primary_data to ensure you only retrieve unique cells.

The best way to run population-scale analyses is using the CELLxGENE Census, which provides versioned single-cell data and supports out-of-core processing for large-scale cross-dataset comparisons without loading everything into memory.

Yes, you can use scanpy and scvi-tools with CELLxGENE Census data. The system supports integration with these tools for in-memory analyses and enables end-to-end machine learning experiments using PyTorch for out-of-core workflows.

To retrieve cell type metadata, you programmatically query the census_info and census_data modules. This allows rapid exploration of standardized dataset metadata and expression matrices across the entire versioned Census.

Filtering for unique cells by is_primary_data is required when querying the CELLxGENE Census. This ensures your cross-tissue analyses and dataset integration workflows avoid duplicate data across the millions of indexed cells.