🌏 pangeapy

pangeapy is a Python package designed for automated cell type annotation and metadata prediction using the PANGEA reference atlas. It enables researchers to easily label single-cell transcriptomic data and predict higher-level phenotypic contexts such as organ or disease state, with minimal preprocessing.

PANGEA reference model can be accessed in two ways:

1. Web interface — Perform cell annotation and explore PANGEA
2. Python package — Integrate annotation and prediction into reproducible pipeline

1) Web interface

You can try cell annotation and explore the PANGEA reference atlas directly at https://pangeapy.org/. (No installation required)

What you can do

1. Annotate cells — Upload .csv or .csv.gz gene expression files and automatically assign cell labels using PANGEA models
2. Explore cell distributions — Visualize organ- and tissue-level patterns
3. Explore gene expression — Plot expression of multiple genes across all cell types

2) Python package

Installation

(1) Create a Conda environment

conda create -n pangea python=3.11
conda activate pangea

(2) Install via GitHub

pip install --user git+https://github.com/srkim727/pangeapy.git

Quick start:

from pangeapy import CellAnnotator, MetaAnnotator

pred = CellAnnotator().annotate(adata)
meta = MetaAnnotator().annotate(pred)

Tutorials

Task	Notebook
01 Cell Annotation	01_vignette_cell_annotation.ipynb
02 Metadata Annotation	02_vignette_meta_annotation.ipynb
03 Missing Cell Type Detection	03_vignette_identifying_missing_cells.ipynb
04 Building Plus One Model	04_vignette_plus_one_model.ipynb

Requirements

celltypist scipy parmap

Citation

Kim, unpublished