cellxgene-census

To query single-cell gene expression data, you can programmatically retrieve records from the CELLxGENE Census by applying filters based on cell type, tissue, disease, and specific gene expression metadata.

Yes, single-cell genomics data retrieved from the Census can be seamlessly integrated into PyTorch and scanpy workflows, directly facilitating machine learning model training and downstream expression analysis.

The best way to explore metadata across millions of human and mouse cells is to programmatically query the CELLxGENE Census, which allows you to discover available datasets, cell types, tissues, and diseases.

Yes, the Census supports both in-memory and out-of-core processing options, enabling efficient data retrieval and analysis even when handling large-scale single-cell genomics datasets that exceed available memory.

Programmatic data queries enable efficient, reproducible filtering across millions of cells by tissue and disease states, facilitating large-scale expression analysis and biomarker identification that manual exploration cannot handle.

Retrieval is limited to the curated datasets within the CELLxGENE Census, meaning analysis depends on the availability of specific cell types, tissues, and diseases present in the current human and mouse collections.