Installation (PyG framework)

Software Dependencies

numpy
opencv-python
pandas
PyYAML
scanpy
scikit-learn
scipy
timm
torch-geometric
torch==1.13.0+cu117
torchaudio==0.13.0+cu117
torchvision==0.14.0+cu117

Installation

After installing the required packages, simply run the following command to install the S3RL package:

pip install S3RL --index-url https://pypi.org/simple

The code is tested with Python 3.7.12 and PyTorch 1.13.0 on a single NVIDIA GeForce RTX 3090 GPU. If you encounter any issues, please check the compatibility of the packages in requirements.txt with your Python version. Additionally, different versions of libraries and different GPU devices may lead to varied outcomes, so to reproduce our results, please use the same versions and hardware configuration as specified.

Data Preparation

Download the datasets and place them in the Data directory, ensuring the directory structure appears as follows:

Data
├── DLPFCs
│   ├── 151673
│   ├── 151674
│   └── ...
├── Nanostring
├── Human_Breast_Cancer
└── Mouse_Brain_Anterior

All datasets can be downloaded from the following links:

• DLPFC

• Nanostring

• Human Breast Cancer

• Mouse Brain Anterior

• Human Lung Cancer

• Mouse Hippocampus Starmap

• Soybean

We also provide the processed datasets for the four datasets used in our experiments, which can be accessed via the link processed datasets.

Additionally, we offer the SimCLR code used for extracting semantic features from the images, available at the link SimCLR. The extracted features are available at the link semantic features.

Finally, the configurations used in our experiments are available at the link configuration, enabling the reproduction of the results reported in the paper.