Spage2vec: Unsupervised representation of localized spatial gene expression signatures

This repository contains a collection of python notebooks for reproducing analyses and results from the original publication [1]. The notebooks folder contains code for:

• Generate spatial gene expression network from in situ transcriptomic data and train an unsupervised graph representation model for producing a node embedding (spage2vec_*.ipynb)
• Visualize and cluster the learned representations in subcelluar funcional domain (*_embedding.ipynb)

System requirement

The sorce code presented in this repository has been developed and tested on a Linux machine running Ubuntu 16.04 operating system with 64GB RAM, Intel(R) Core(TM) i7-5930K CPU @ 3.50GHz cpu, and nvidia TITAN X gpu.

Python Library Requirements

The following python packages are required for running the notebooks:

• numpy==1.17.2
• tensorflow==1.12.0
• tensorboard==1.12.2
• networkx==2.4
• pandas==0.25.2
• matplotlib==3.0.3
• stellargraph==0.8.1
• scipy==1.3.1
• scikit-learn>=0.21.3
• tqdm==4.36.1
• umap-learn==0.3.10
• scanpy==1.4.4
• leidenalg==0.7.0
• seaborn==0.9.0
• h5py==2.10.0
• loompy==3.0.6

Data Download

Spatial gene expression data for the analyzed assays can be downloaded at: https://doi.org/10.5281/zenodo.3897401. Please extract the content of the zipped archive in this repository local folder before running the notebooks.

Citation

[1] Partel, G., and Wählby C. Spage2vec: Unsupervised detection of spatial gene expression constellations. BioRxiv, https://doi.org/10.1101/2020.02.12.945345, (2019).