Scanpy

Warnings

• breaking Scanpy 1.12.0 removed support for Python versions older than 3.12 and now requires anndata>=0.10. Users on older Python environments will need to upgrade. Scanpy 1.10.4 also removed Python 3.9 support.
  Fix: Upgrade Python to 3.12 or newer. Ensure `anndata` is updated to at least version 0.10. `pip install 'scanpy>=1.12.0'` will manage `anndata` dependencies appropriately.
• gotcha Directly using Scanpy's internal (non-public) APIs is not officially supported and may lead to breaking changes in minor or patch releases. Stick to the documented public API.
  Fix: Refer to the official API documentation for supported functions and classes. If a desired feature is not in the public API, consider opening an issue on the Scanpy GitHub repository.
• gotcha Reproducibility of `sc.tl.leiden` clustering results, even with `random_state` set, has been reported to be inconsistent between different Scanpy minor versions (e.g., 1.9.3 vs 1.10.4). This may be due to changes in underlying dependencies like `numpy` or `sc.pp.neighbors`.
  Fix: For critical reproducibility, tightly pin all major dependencies (Scanpy, anndata, numpy, scikit-learn, leidenalg, igraph). Be aware that perfect bit-for-bit reproducibility across all environments might be challenging. Document your full environment using `pip freeze` or similar.
• deprecated The `scanpy.__version__` attribute is deprecated. Use `scanpy.version()` instead.
  Fix: Replace `sc.__version__` with `sc.version()` to retrieve the library version.
• deprecated Some functions within `scanpy.pp` (preprocessing module) are raising `FutureWarning` due to upcoming changes.
  Fix: Pay attention to `FutureWarning` messages in your console output and adapt your code to the suggested new patterns or parameters as indicated in the warnings or release notes.

Install

• pip install scanpy Basic installation
• pip install 'scanpy[leiden]' Recommended for clustering (includes igraph and leidenalg)

Imports

• scanpy

  import scanpy as sc

  The conventional alias for Scanpy.
• AnnData

  import anndata as ad

  While AnnData is a core concept, it's typically imported as 'ad' separately, or accessed via Scanpy functions returning AnnData objects.

Quickstart

This quickstart demonstrates a typical single-cell RNA sequencing (scRNA-seq) analysis workflow using Scanpy. It covers loading a dataset, essential preprocessing steps (filtering, normalization, log-transformation, highly variable gene selection, regression, scaling), dimensionality reduction (PCA, UMAP), and clustering (Leiden). Finally, it visualizes the UMAP embedding colored by cluster and QC metrics. The `pbmc3k` dataset is used as a readily available example.

import scanpy as sc
import matplotlib.pyplot as plt

# Set verbosity for Scanpy (0: errors, 1: warnings, 2: info, 3: hints)
sc.settings.verbosity = 3

# Load a sample dataset (e.g., pbmc3k from 10x Genomics)
# This downloads the data if not present in the datasetdir
adata = sc.datasets.pbmc3k()

# Basic preprocessing pipeline
sc.pp.filter_cells(adata, min_genes=200)
sc.pp.filter_genes(adata, min_cells=3)
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
sc.pp.highly_variable_genes(adata, min_mean=0.0125, max_mean=3, min_disp=0.5)
adata = adata[:, adata.var.highly_variable]
sc.pp.regress_out(adata, ['total_counts', 'pct_counts_mt'])
sc.pp.scale(adata, max_value=10)

# Dimensionality reduction and clustering
sc.pp.pca(adata)
sc.pp.neighbors(adata, n_neighbors=10, n_pcs=40)
sc.tl.umap(adata)
sc.tl.leiden(adata)

# Visualization
sc.pl.umap(adata, color=['leiden', 'n_genes_by_counts', 'total_counts'], show=False)
plt.tight_layout()
plt.show()