Can I use this with Scanpy?

Yes, the skill includes patterns for using the get_anndata() function to retrieve Census data directly into Scanpy-compatible objects for standard analysis workflows.

Does this skill handle datasets that are too large for memory?

Yes, it includes patterns for out-of-core processing using axis_query and iterative table processing to handle data that exceeds available RAM.

Why should I filter for is_primary_data?

Filtering for is_primary_data == True is a best practice to ensure you avoid counting duplicate cells that might be present across different datasets within the Census.

What is the CZ CELLxGENE Census?

It is a comprehensive, versioned collection of standardized single-cell genomics data providing access to over 61 million cells from human and mouse datasets.

How does this skill help with machine learning?

This skill provides specific patterns for using the PyTorch experimental dataloader, allowing you to train models directly on Census data without loading the entire dataset into RAM.

CZ CELLxGENE Census

Name: CZ CELLxGENE Census
Author: lifangda

bylifangda

•

Data Science & ML

Facilitates programmatic access and analysis of the CZ CELLxGENE Census database containing over 61 million single-cell genomics records.

About

This skill empowers researchers and bioinformatics engineers to efficiently query, filter, and analyze massive single-cell datasets from the CZ CELLxGENE Discovery platform. It streamlines the process of retrieving gene expression matrices based on specific cell types, tissues, or diseases, while offering seamless integration with popular analysis frameworks like Scanpy and PyTorch for machine learning tasks. Whether performing population-scale cross-dataset analyses or training specialized biological models, this skill provides the patterns and best practices needed to handle million-cell datasets with optimized memory management and reproducibility.

Key Features

16 GitHub stars
Efficient querying of 61M+ standardized single-cell genomics records
Metadata filtering by cell type, tissue, disease, and organism (human/mouse)
Memory-efficient out-of-core processing for large-scale queries
Experimental PyTorch support for training ML models on single-cell data
Seamless integration with Scanpy for downstream analysis and visualization

Use Cases

Creating custom sub-sampled datasets for specialized bioinformatics research
Training machine learning models on large-scale single-cell atlas data
Performing cross-tissue or cross-disease comparative gene expression analysis

About

Key Features

16 GitHub stars
Efficient querying of 61M+ standardized single-cell genomics records
Metadata filtering by cell type, tissue, disease, and organism (human/mouse)
Memory-efficient out-of-core processing for large-scale queries
Experimental PyTorch support for training ML models on single-cell data
Seamless integration with Scanpy for downstream analysis and visualization

Use Cases

Creating custom sub-sampled datasets for specialized bioinformatics research
Training machine learning models on large-scale single-cell atlas data
Performing cross-tissue or cross-disease comparative gene expression analysis