Spatial gene expression at single-cell resolution from histology using deep learning with GHIST

For more details, please refer to our paper.

GHIST is a deep learning-based framework that predicts spatial gene expression at single-cell resolution from histology (H&E-stained) images by leveraging subcellular spatial transcriptomics and synergistic relationships between multiple layers of biological information.

Installation

Note: A GPU with 24GB VRAM is strongly recommended for the deep learning component. We ran GHIST on a Linux system with a 24GB NVIDIA GeForce RTX 4090 GPU, Intel(R) Core(TM) i9-13900F CPU @ 5.60GHz with 32 threads, and 32GB RAM.

1. Clone repository:

git clone https://github.com/SydneyBioX/GHIST.git

2. Create virtual environment:

conda create --name ghist python=3.10

3. Activate virtual environment:

conda activate ghist

4. Install dependencies:

cd GHIST
pip install -r requirements.txt

Please install the stainlib package from https://github.com/sebastianffx/stainlib

Typically installation is expected to be completed within a few minutes.

Tutorials

Please check out the examples in tutorials for key use cases of GHIST, including:

• Data pre-processing
• Training and validation
• Prediction on data without ground truth

Figures

Code for creating the output figures may be found in https://github.com/SydneyBioX/GHIST_figure

Citation

If GHIST has assisted you with your work, please kindly cite our paper:

• https://www.biorxiv.org/content/10.1101/2024.07.02.601790v1