SpacGPA: Spatial and single-cell Gene Program Analysis

SpacGPA is a GPU-accelerated toolkit that annotates spatial transcriptomes through de novo interpretable gene programs. It builds co-expression networks via a Gaussian graphical model (GGM), identifies programs with a modified Markov Clustering (MCL) algorithm, performs ontology-based enrichment (Gene Ontology (GO) / Mammalian Phenotype (MP)), and applies programs to spatial analyses such as detection of SVGs, spatial domain annotation, and label integration.

---

This website hosts the installation guide, in-depth workflows, and full API reference.

Quick Links

• Installation guide → Installation

• In-depth workflows → Tutorials

• Full API reference → api/index

Table of Contents

Getting Started

• Installation

Tutorials

• Tutorials
  • Tutorial 1: Stereo-Seq: MOSTA E16.5 E2S5
  • Tutorial 2: Visium HD: Human Tonsil
  • Tutorial 3: Xenium Prime 5K: Mouse Brain
  • Tutorial 4: MERFISH: Mouse Brain
  • Tutorial 5: Visium HD: Soybean Seed
  • Tutorial 6: Atera: Breast Tumor

---

Citation

If you use SpacGPA in your work, please cite: Xu Y, Chen L, Ma S. SpacGPA: annotating spatial transcriptomes through de novo interpretable gene programs. bioRxiv (2025). https://doi.org/10.1101/2025.10.01.679918

Contact

• GitHub : https://github.com/MaShisongLab/SpacGPA

• Issues : Please open a ticket on GitHub Issues for bugs or feature requests

• E-mail : sma@ustc.edu.cn