The pages that run code read shared datasets by file name, such as
read_cube("simulated_bulk.cube.zarr"). A script in the repository,
docs/datasets.py,
makes them deterministically into a cache outside the repository, downloads what they
need with checksums, and checks each against the fingerprint of every cube and the
digest of every other file it records. From the repository root:
uv run --group docs python docs/datasets.py list
uv run --group docs python docs/datasets.py prepare simulated_pbmc
uv run --group docs python docs/datasets.py verifyThe cache is CYTEPOOL_DOCS_DATA, else cytepool-docs under XDG_CACHE_HOME or
~/.cache. Copy or link the files a page names into your working directory to run its
code.
| Dataset | What it holds | Source | Licence | Cite |
|---|---|---|---|---|
simulated_pbmc | 12 donors, each sampled before and after a stimulation; about 20,000 cells of six cell types over 2,000 genes, with planted changes in every cell type, a response program shared by all, a rise in the share of monocytes, and dendritic cells too rare in some samples; the cube of all five layers and the cells with the true effects | simulated by the script, seed 20261026 | part of CytePool, BSD 3-Clause | |
simulated_bulk | 3 treatments x 4 replicates in two batches, 10,000 genes; drug A changes 500 genes, drug B 300; one noisier sample; the cube and the true effects | simulated by the script, seed 20261027 | part of CytePool, BSD 3-Clause | |
kang2018 | Peripheral blood cells of 8 lupus patients, each split into an interferon-beta stimulated and an untreated culture: 24,673 cells pooled into 16 samples x 8 cell types x 35,635 genes (counts and detecting cells); a 77 MB download | GEO GSE96583, batch 2 | GEO’s terms: NCBI places no restrictions on the use of GEO data, though submitters may claim rights in theirs (disclaimer) | Kang et al. (2018) |
hallmark_interferon_alpha | The 97 genes of the Hallmark interferon-alpha response set (M5911, MSigDB v2026.2.Hs) | MSigDB | Creative Commons Attribution 4.0 (terms) | Liberzon et al. (2015) |
The downloads of kang2018 are the files that the validation script
validation/interferon_stimulated_pbmcs.py downloads, with the same checksums and in
the same directory, so a machine fetches them once; its cube is the one that script
tests. A notebook page names its datasets in its metadata; make -C docs execute
prepares them, links them into the directory the page runs in, and records the
fingerprint of each file it was given, which the build checks against the registry.
- Kang, H. M., Subramaniam, M., Targ, S., Nguyen, M., Maliskova, L., McCarthy, E., Wan, E., Wong, S., Byrnes, L., Lanata, C. M., Gate, R. E., Mostafavi, S., Marson, A., Zaitlen, N., Criswell, L. A., & Ye, C. J. (2018). Multiplexed droplet single-cell RNA-sequencing using natural genetic variation. Nature Biotechnology, 36(1), 89–94. 10.1038/nbt.4042
- Liberzon, A., Birger, C., Thorvaldsdóttir, H., Ghandi, M., Mesirov, J. P., & Tamayo, P. (2015). The Molecular Signatures Database hallmark gene set collection. Cell Systems, 1(6), 417–425. 10.1016/j.cels.2015.12.004