CyteBase

CyteBase is a searchable catalog of public single-cell datasets stored as cloud-hosted CyteArc DataStores. Search by study, tissue, disease, organism, or cell type, inspect a dataset's metadata and published embeddings, then open it read-only or mount it for your own analysis. Counts stay remote on Hugging Face, and every result you compute is saved in a store you control. Public access needs no credentials.

Each release is a CELLxGENE Discover snapshot converted to CyteArc stores and published in its own bucket. The current release, 2610, holds the primary RNA datasets of 376 registration-ready collections from the snapshot taken on 2026-10-10.

Install

Python 3.13 or newer.

pip install cytebase

CyteBase installs CyteArc. Install cytearc[extra] as well to plot datasets with ds.plots.

Quick start

import cytebase

catalog = cytebase.Catalog()  # the latest public release, read anonymously
matches = catalog.search("kidney")
print(matches)

# Open one of the matching datasets for read-only exploration.
dataset_id = matches[0]["cytebase_id"]
ds = catalog.open_datastore(dataset_id)

# Create a writable local store for your own analysis, keeping counts remote.
analysis = catalog.mount_datastore(dataset_id, at="kidney_analysis.zarr")

analysis is a CyteArc DataStore: run the CyteArc pipeline on it to save new results locally while count blocks are read over the network as needed. cytebase.releases() lists the public releases, and cytebase.Catalog(version="2610") opens a specific one, so a result can name the exact release it came from.

Read the documentation for catalog filters, SQL queries, embeddings, mounts, and case studies.

Working with CyteArc

CyteBase depends on CyteArc, which opens and analyzes the datasets: read-only DataStores, mounts for your own results, and composite stores that analyze several remote datasets together without copying their counts. CyteArc does not depend on CyteBase. Each dataset is a CyteArc store at a Hugging Face URI, the catalog's zarr_uri, so CyteArc alone can open it too, as the CyteArc documentation shows.

Repository layout

• cytebase/: the SDK published to PyPI.
• cytebase/pipeline/: the Modal pipeline that ingests CELLxGENE datasets, converts and verifies them, and publishes the catalog. It runs from a checkout and is not part of the wheel.
• docs/: the documentation site.
• skills/cytebase/: the agent skill for using CyteBase.
• tests/: the test suite.

Contributors and coding agents start with AGENTS.md.

License

BSD 3-Clause. See LICENSE. Dataset licenses and citations come from their sources; each catalog entry records its citation.