SubCellSpace

A latent space of representing subcellular mRNA localization patterns

Wouters et al. 2026

---

Overview

SubCellSpace encodes subcellular mRNA localization patterns into a compact, interpretable latent space. This repository contains all code needed to reproduce the results and figures of Wouters et al. 2026.

---

Repository Structure

Each module ships with its own dedicated conda environment (see yamls/):

Module	Description
training_subcellspace/	Build training data and train the SubCellSpace model
make_paper_figures/	Reproduce all manuscript analyses and figures
raw_spatial_data_processing/	Process your own dataset into a SubCellSpace-ready AnnData object

Getting Started

Option 1 — Conda (recommended)

Create and activate an environment for the module you want to run. For example, to reproduce the paper figures:

conda env create -n subcellspace -f ./yamls/analysis_figures.yml
conda activate subcellspace

Each subfolder contains a dedicated README with step-by-step instructions.

Option 2 — Docker

A Docker container is also available if you prefer a fully isolated setup:

docker pull woutdavid/subcellspace_jupyter_env
docker run -it --rm -p 8888:8888 -v $(pwd):/workspace woutdavid/subcellspace_jupyter_env

Then open the URL printed in the terminal (e.g. http://127.0.0.1:8888/?token=...) in your browser.

---

Applying SubCellSpace to your own data

Once your environment is set up, this is all you need:

1. Process your dataset using raw_spatial_data_processing/
2. Follow the example notebook: example_usage_of_SubCellSpace_on_your_data.ipynb

No need to retrain the model or reproduce the paper figures.

---

Citation

If you use SubCellSpace in your research, please cite:

@article{wouters2026subcellspace,
  title   = {SubCellSpace: A latent space of subcellular mRNA localization patterns},
  author  = {Wouters et al.},
  journal = {bioRxiv},
  year    = {2026},
  doi     = {10.64898/2026.04.28.720613},
  url     = {https://www.biorxiv.org/content/10.64898/2026.04.28.720613v1}
}

---