How do I handle datasets too large for my RAM?

Use the skill's out-of-core axis_query() method to iterate through expression matrices in chunks using tiledbsoma, allowing you to process millions of cells without memory exhaustion.

Why is the 'is_primary_data' filter important?

Setting is_primary_data to True ensures you are only analyzing unique cell entries, preventing statistical bias from duplicate cells present in overlapping datasets.

Does this skill support both human and mouse data?

Yes, the CZ CELLxGENE Census provides standardized genomic data and metadata for both Homo sapiens and Mus musculus organisms.

Can I use this for training deep learning models?

Absolutely. It includes experimental PyTorch integration that provides specialized dataloaders for streaming genomic data directly into training loops.

How do I ensure my genomic analysis is reproducible?

The skill encourages specifying a specific 'census_version' string when opening the connection, ensuring your code points to a fixed snapshot of the data.

CZ CELLxGENE Census Analysis

Name: CZ CELLxGENE Census Analysis
Author: ricable

byricable

•

Data Science & ML

Queries and analyzes millions of standardized single-cell genomics data points from the CZ CELLxGENE Census for population-scale biological research.

About

This skill provides programmatic access to the CZ CELLxGENE Census, a massive repository containing over 61 million cells from human and mouse studies. It enables bioinformaticians and researchers to efficiently filter data by cell type, tissue, or disease state and retrieve raw gene expression matrices. By bridging the gap between raw genomic data and analysis frameworks, it facilitates large-scale cross-dataset studies, machine learning model training via PyTorch, and seamless integration with the Scanpy ecosystem, making it essential for high-throughput single-cell discovery.

Key Features

Seamless conversion to AnnData objects for Scanpy bioinformatics workflows
Native integration with PyTorch for training genomic machine learning models
Programmatic access to 61M+ standardized single-cell genomic profiles
Out-of-core processing for handling datasets that exceed available RAM
Granular metadata filtering by tissue, disease, cell type, and donor
1 GitHub stars

Use Cases

Extracting specific cell populations for downstream differential expression testing
Training large-scale foundation models on massive single-cell expression datasets
Performing cross-tissue comparative analysis of gene expression in specific diseases

About

Key Features

Seamless conversion to AnnData objects for Scanpy bioinformatics workflows
Native integration with PyTorch for training genomic machine learning models
Programmatic access to 61M+ standardized single-cell genomic profiles
Out-of-core processing for handling datasets that exceed available RAM
Granular metadata filtering by tissue, disease, cell type, and donor
1 GitHub stars

Use Cases

Extracting specific cell populations for downstream differential expression testing
Training large-scale foundation models on massive single-cell expression datasets
Performing cross-tissue comparative analysis of gene expression in specific diseases