cellxgene-census

To query single-cell RNA-seq data, you can programmatically access standardized genomics data to filter specific cell types like 'B cell' across tissues like 'lung' at population scale. This provides efficient access to cross-dataset expression data for comparative analysis.

Population-scale single-cell data analysis involves computing statistics and training machine learning models across millions of cells. You need this approach when standard analysis tools cannot handle querying massive datasets like the 61M+ cell CELLxGENE Census.

Yes, cellxgene-census integrates seamlessly with scanpy and PyTorch for advanced single-cell data analysis. You can use these dependencies alongside scvi-tools to train machine learning models and perform complex computations on the queried expression data.

You do not need scvi-tools for basic querying, as the required cellxgene_census library handles data access independently. However, scvi-tools is an optional dependency that provides advanced capabilities for integrating and analyzing cross-dataset single-cell genomics data.

The best way to access large-scale single-cell expression data is using a programmatic interface to a versioned collection like the CELLxGENE Census. This allows you to efficiently query millions of cells and directly train machine learning models without manual data aggregation.

When training machine learning models on 61 million cells, limitations depend on your available compute resources and PyTorch configuration rather than the data query itself. The Skill efficiently retrieves the single-cell data, but model training scale is constrained by local hardware.