This repository contains the source code for Truffle (Trajectory Inference via Multi-commodity Flow with Node Constraints).

Processed and clustered data can be downloaded from Google Drive. Place these files inside the data folder. To reduce and cluster the data, we used the grinch library. The config files used for grinch can be found in the conf folder.

To run the code in this repository, grinch needs to be installed as follows:

git clone https://github.com/euxhenh/grinch
cd grinch
pip install -e .

To run Truffle, see the notebook notebooks/multi-commodity_flow.ipynb.

• pyomo implementation of multi-commodity flow is under src/mc_flow.py
• Truffle's main class is under src/truffle.py
• R scripts used to run Tempora and psupertime can be found in src/R

If you wish to perform your own clustering, update the .obs['leiden'] key in the AnnData object. Patient visits are under .obs['visit']. If you wish to use grinch for clustering, run the following from the root of Truffle's directory:

python /path-to-grinch-folder/src/grinch/main.py conf/config-file-you-wish-to-use.yaml

This will start the grinch pipeline and apply all the steps specified in the config. Feel free to update parameters.