CZ CELLxGENE Census — анализ single-cell транскриптомики

★ 7.0 · research

cellxgene-census is a Claude Code skill that provides programmatic access to the CZ CELLxGENE Census for querying versioned public single-cell and spatial transcriptomics data. The 2025-11-08 stable LTS release contains over 217 million total cells across 1,845 datasets, covering human, mouse, marmoset, rhesus macaque, and chimpanzee. The skill delivers eight core workflow patterns: opening the Census with a pinned version for reproducibility, exploring available datasets and summary counts, querying gene expression into AnnData, out-of-core processing for large slices, PyTorch model training via TileDB-SOMA-ML, accessing spatial assays, integrating with Scanpy, and handling multi-dataset batch correction. It suits population-scale analyses by cell type, tissue, disease, or assay, as well as reference atlas comparisons and machine learning on single-cell data. Requires Python ≥3.10,<3.13; no authentication is needed for public Census data.