A self-contained tutorial that opens the CZ CELLxGENE Discover Census, pulls a stratified sample of 3,600 adult human retina nuclei from one published dataset into an AnnData object, runs the standard scanpy workflow (QC, normalization, feature selection, PCA, neighbors, UMAP, Leiden, marker genes), and compares the clusters with the Census cell-type labels to separate real sub-structure from batch effects and ambient RNA.
Notebook: census_retina_scanpy.ipynb
Listed on scverse.org/learn.
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
jupyter lab census_retina_scanpy.ipynb
Requirements: Python 3.10 or later, a network connection, about 2 GB of free memory. The notebook downloads roughly 100 MB from the Census and runs in under five minutes on a laptop. All data are fetched inside the notebook; nothing needs to be downloaded in advance.
The Census release is pinned to 2025-11-08 (a long-term supported release) so
the sampled cells and the results are stable over time.
Orozco LD et al., Integration of eQTL and a Single-Cell Atlas in the Human Eye
Identifies Causal Genes for Age-Related Macular Degeneration, Cell Reports
(2020), https://doi.org/10.1016/j.celrep.2019.12.082, as distributed by CZ
CELLxGENE Discover (dataset e6dad530-418b-47f9-af6e-472e56a7b314).
Aaron Kanzer (Scigantic). Questions and corrections: open an issue here or reach me on the scverse Zulip.
BSD 3-Clause, see LICENSE.