Python's workhorse for million-cell biology datasets

What it does

Scanpy is a Python toolkit for analyzing single-cell gene expression data, covering the full pipeline from preprocessing and visualization to clustering, trajectory inference, and differential expression testing. It is built jointly with anndata, its companion data structure for annotated matrices. The project claims efficient handling of datasets exceeding one million cells.

The interesting bit

The project is now threading experimental Dask support through many functions, letting researchers work with datasets too large for memory without leaving the familiar Python API. That is the boring part that matters: single-cell datasets have grown faster than RAM budgets, and Scanpy is trying to bridge that gap without forcing users into an entirely different framework.

Key highlights

• Part of the broader scverse ecosystem, with NumFOCUS fiscal sponsorship
• Public API is documented and stability-guaranteed; internal APIs are explicitly not
• Active community support via Discourse and Zulip chat
• Published in Genome Biology (2018) with a follow-up scverse paper in Nature Biotechnology (2023)

Caveats

• Dask integration is marked experimental; not all functions support out-of-core execution
• The README warns that internal APIs are unstable and officially unsupported, though they acknowledge many users rely on them anyway

Verdict

Computational biologists and data scientists working with single-cell transcriptomics should already know this tool; it is effectively standard infrastructure in the field. General Python developers without a biological use case can safely look elsewhere.