SEAL: Spatially-resolved Embedding Analysis with Linked Imaging Data

Simon Warchol¹'², Grace Guo¹'², Johannes Knittel¹, Dan Freeman², Usha Bhalla¹, Jeremy Muhlich², Peter K. Sorger², and Hanspeter Pfister¹

¹ Harvard John A. Paulson School of Engineering and Applied Sciences
² Laboratory of Systems Pharmacology, Harvard Medical School

Paper

📄 bioRxiv Preprint

Demo Datasets

Lung Adenocarcinoma Specimen

From a larger tissue microarray (TMA), imaged using CyCIF with three cycles. Each cycle consists of six four-channel image tiles.

📔 Open in Google Colab

Sloan Digital Sky Survey

Multi-spectral SDSS imaging of galaxies NGC 450 and UGC 807, capturing structural and spectral features across five filters.

Installation

pip install seal

or with uv:

uv add seal

Development

We recommend using uv for development. It will automatically manage virtual environments and dependencies for you.

uv run jupyter lab example.ipynb

Alternatively, create and manage your own virtual environment:

python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
jupyter lab example.ipynb

The widget front-end code bundles it's JavaScript dependencies. After setting up Python, make sure to install these dependencies locally:

npm install

While developing, you can run the following in a separate terminal to automatically rebuild JavaScript as you make changes:

npm run dev

Open example.ipynb in JupyterLab, VS Code, or your favorite editor to start developing. Changes made in js/ will be reflected in the notebook.