We recommend the usage of a conda environment for running the experiments.
The used conda environment can be found at environment.yml and used with the command (The created environment is called hypegraph):
conda env create -f environment.yml
Once the environment is created you just need to activate it by:
conda activate hypergraph
The data can be downloaded on the following link: https://sea-ad-spatial-transcriptomics.s3.amazonaws.com/index.html#middle-temporal-gyrus/all_donors-h5ad/
You can then extract your zip file into the data/raw folder.
After all the data was extracted into the raw files, you can use the notebook found at : notebooks/create_patients_section_files.ipynb to process the raw data and divide the dataset from patients level to section level data.
This notebook will create all the section data in the folder: data/interim/section_data
Once the data is on the correct location you can simple run python3 main.py
The resulting wavelets features will be stored on data/processed/wavelet_features
We evaluated performance using vendi score and linear probing. Each one of those can be computed on the following scripts:
python3 vendi.py
python3 linear_probing.py