Single-Cell Packages

Purpose

This page is for collecting packages encountered while learning and doing single-cell analysis.

Keep it practical: package name, ecosystem, task, and when it becomes relevant.

Packages

Package	Ecosystem	Task	Notes
Seurat	R	General workflow	Common entry point for scRNA-seq case analysis
scRNAseq	R / Bioconductor	Public datasets	Curated scRNA-seq datasets returned as SingleCellExperiment objects
patchwork	R	Plot composition	Combine ggplot figures into one layout
SingleCellExperiment	R / Bioconductor	Data object	Core Bioconductor container
zellkonverter	R / Bioconductor	h5ad import/export	Bridges AnnData and SingleCellExperiment
Scanpy	Python	General workflow	Main AnnData-based analysis framework
AnnData	Python	Data object	Common `.h5ad` object format

Package Notes

Seurat

Seurat is the main R package I will encounter for practical scRNA-seq analysis. It is often used as a full workflow framework rather than a single-purpose tool.

Typical uses:

create a single-cell object from count matrices
store cell metadata, gene metadata, assays, reductions, graphs, and cluster labels
run QC, normalization, feature selection, PCA, UMAP, clustering, marker detection, and integration
handle common multimodal extensions such as CITE-seq, spatial data, and multi-assay objects

What to remember:

Seurat is convenient because many steps live in one ecosystem.
A Seurat object is not just an expression matrix; it also stores metadata and analysis results.
For condition-level differential expression, do not blindly use cell-level marker tests as final evidence. For multi-sample disease comparisons, pseudobulk is usually the safer direction.
When importing from .h5ad, check whether raw counts, normalized data, layers, reductions, and metadata survived conversion.

patchwork

patchwork is an R package for combining multiple ggplot2 plots into one figure layout.

In single-cell work, it is useful when comparing related plots side by side, such as QC plots, UMAPs, feature plots, dot plots, or condition-specific panels.

Common pattern:

p1 + p2
p1 / p2
(p1 + p2) / p3

What to remember:

+ places plots side by side.
/ stacks plots vertically.
Parentheses control the layout.
It is mainly for figure assembly, not data analysis.

scRNAseq

scRNAseq is a Bioconductor data package that provides curated public single-cell RNA-seq datasets. It does not implement an analysis workflow; its role is to supply ready-to-use example data for learning, method testing, and reproducible demonstrations.

Common datasets include:

ZeiselBrainData() — mouse brain cells
MacoskoRetinaData() — mouse retina cells
PaulHSCData() — hematopoietic stem and progenitor cells

library(scRNAseq)

sce <- ZeiselBrainData()
sce

The returned object is usually a SingleCellExperiment, so it fits directly into the Bioconductor ecosystem and can also be converted for use with Seurat when needed.

What to remember:

this is a data package, not an analysis package
check the documentation for the experiment, organism, assay, and available metadata before using a dataset
downloading a dataset may require network access and local caching
example datasets are useful for testing code, but they do not replace data appropriate to the actual research question

Open Questions

Should data objects have separate pages?
Should this stay as a package list or become a workflow map?
Should R and Python ecosystems be separated later?