Run scripts from the root directory of this repo.

Code for training the encoder model are contained in the learn subfolder with training scripts in learn/scripts.

Code for traversal and alignment are contained in the traversal subfolder with example scripts in traversal/scripts.

All code for generating the figures presented in the paper are contained within the paper subfolder. To examine the methods used to generate a figure, check out it's corresponding notebook.

Colon and stomach datasets are available upon reasonable request to the corresponding author.

Data should be organized as follows:

.
├── colon (organ)
│   ├── CD (disease or patient)
│   │   ├── A (slide)
│   │   │   ├── A.tif
│   │   │   └── outs (spaceranger output)
│   │   │       ├── filtered_feature_bc_matrix.h5
│   │   │       └── spatial
│   │   │           ├── scalefactors_json.json
│   │   │           ├── tissue_hires_image.png
│   │   │           └── tissue_positions_list.csv
│   │   └── B
│   │       ├── B.tif
│   │       └── outs
│   │           └── [...]
│   └── UC
│       └── [...]
└── dlpfc
    └── [...]

GitHub classifies this repository as 99% Jupyter Notebook because the paper figures are saved within each notebook, thus drastically inflating the file size of each notebook. The bulk of the methods are however written in plain python files.