Chimeric Antigen Receptor Optimal Transport (CAROT)

The conditional Monge Gap applied the single cell RNA sequencing data of Chimeric Antigen Receptor T cells. Extension of the Conditional Monge Gap, to include CAR specific dataloaders, embeddings and trainers. Additionally notebooks contains Notebooks for generating the figures of the preprint and additional analyses. In the configs and scripts directories are all scripts to replicate the experiments from this preprint.

Development setup & installation

We use poetry as package manager and tested the code in Python 3.10.

pip install poetry # into your base env
git clone git@github.com:AI4SCR/car-conditional-monge.git
cd car-conditional-monge
poetry install -v

If the installation was successful, activate the env interatively via poetry shell.

Example usage

You can find example config in tests/configs/ for the unconditional and the conditional setting. To train a conditional monge model:

from carot.datasets.conditional_loader import ConditionalDataModule
from carot.trainers.conditional_monge_trainer import ConditionalMongeTrainer
from cmonge.utils import load_config


config_path = Path("tests/configs/conditional_synthetic.yml")
config = load_config(config_path)

datamodule = ConditionalDataModule(config.data, config.condition)

logger_path = Path(config.logger_path)

datamodule = ConditionalDataModule(config.data, config.condition)
trainer = ConditionalMongeTrainer(jobid=1, logger_path=logger_path, config=config.model, datamodule=datamodule)

trainer.train(datamodule)
trainer.evaluate(datamodule)

Citation

If you find this work useful, please cite:

@inproceedings{driessen2024modeling,
  title={Modeling CAR Response at the Single-Cell Level Using Conditional OT},
  author={Driessen, Alice and Born, Jannis and Rueda, Roc{\'\i}o Castellanos and Reddy, Sai T and Rapsomaniki, Marianna},
  year={2024},
  booktitle={NeurIPS 2024 Workshop on AI for New Drug Modalities},
  note={Spotlight talk}
}