SpatialDIVA 0.1b0

Multi-modal Disentanglement of Spatial Transcriptomics and Histopathology Imaging

SpatialDIVA - disentangling spatial transcriptomics and histopathology data

This repository contains code for the SpatialDIVA method, associated preprocessing, and evaluations performed in the manuscript - "Multi-modal disentanglement of spatial transcriptomics and histopathology imaging".

Table of Contents

• Installation
• Datasets
• Usage
• Tutorials
• Preprocessing
• Paper evaluation code
• Citation

Installation

Coming soon!

Processed datasets

Valdeolivas et al. - colorectal cancer (https://www.nature.com/articles/s41698-023-00488-4)

This dataset can be downloaded from Figshare at https://figshare.com/s/e12b576b1b05cb1ab77d. After downloading, please move the data to the following directory:

mkdir spatialdiva/data
mv valdeolivas_processed spatialdiva/data

Zhou et al. - pancreatic ductal adenocarcinoma (https://www.nature.com/articles/s41588-022-01157-1)

Coming soon!

Usage

Coming soon!

Tutorials

The following notebooks offer more in-depth tutorials on how to use the SpatialDIVA model for relevant analyses of histopathology and spatial transcriptomics data:

1. Factor covariance analysis with SpatialDIVA - spatialdiva/tutorials/01_colorectal_cancer_spdiva_analysis.ipynb

2. Conditional generation analysis with SpatialDIVA - Coming soon!

3. Tumor annotation and subtyping with SpatialDIVA - Coming soon!

Preprocessing

The preprocessed data for Valdeolivas et al. (colorectal cancer) and Zhou et al. (pancreatic cancer) contains spot-aligned features for histopathology imaging extracted using the UNI foundation model (https://github.com/mahmoodlab/UNI).

Code for preprocessing in-house datasets in a similar manner, as well as environment and installation information is available in the spatialdiva/preprocessing directory.

Paper evaluation code

Coming soon!

Citation

TBD