Tutorials#
Notebooks are pre-executed. The HTML on Read the Docs already has the tables and figures.
Notebook |
What it does |
|---|---|
Public PBMC 3k. Load, QC, default scFair, Leiden / UMAP. No labels. |
|
Labeled PBMC 10k. Exact top-2000 vs top-2000 + append, same Leiden recipe. |
PBMC 3k downloads its own data.
PBMC 10k: top-2000 vs top-2000 plus append reads this repository-relative file:
examples/data/pbmc_10k_v3_labeled.h5ad
That object is not on PyPI. Public 10x PBMC 10k counts are at 10x Genomics. Any AnnData with raw counts and a cell-type column can follow the same comparison. Runtime with data on disk: a few minutes for two HVG → Leiden → UMAP passes on ~10k cells.
pip install scfair
jupyter lab docs/tutorials/
PBMC 3k
QC → default HVG → PCA / Leiden / UMAP, no gold labels.
PBMC 10k
Majority-vote clustering error, top-2000 vs top-2000 + append.