Opponent Shaping for Antibody Development

This is our implementation of the paper on Opponent Shaping for Antibody Development.

It consists of a few elements:

• JAX-accelerated binding simulation, using the Absolut! framework as a base - binding/*
• Simulated viral escape via evolution - gen_alg_basic.py and shaping_func.py
• Optimisation of antibody shapers - main_shaping_process.py

Installation

git clone https://github.com/olakalisz/antibody-shapers.git
cd antibody-shapers
pip install .

This should install the necessary requirements too.

Usage

To get started first you'll need to download the Dengue Antigen data that we used throughout our experiments. We made a GDrive folder with all the relevalnt files. Download the files and move them to the ./data directory in the repository.

You can run an example simluated viral escape to a random antibody in a following way:

python src/antibody_shapers/shaping_funcs.py

This should create a ./results/test_run_results.pkl file. You can investigate the outputs by loading:

import pickle as pkl
with open(f"./results/test_run_results.pkl", "rb") as f:
        test = pkl.load(f)

As an example test["ag_performances"] contains the viral fitness values for a full 100 horizon steps escape trajectory.

Our code also allows for an easy calculation of binding, simulating viral escape and optimising of antibody shapers. More documnetation on how to run all these things coming soon!

Cite our work

If you use Opponent Shaping for Antibody Development in your research, please cite it as follows:

@article{towerskalisz2024OSforantibodies,
  title={Opponent Shaping for Antibody Development},
  author={Sebastian Towers and Aleksandra Kalisz and Philippe A. Robert and Alicia Higueruelo and Francesca Vianello and Ming-Han Chloe Tsai and Harrison Steel and Jakob N. Foerster},
  journal={arXiv preprint arXiv:2409.10588},
  year={2024}
}

License

antibody_shapers was created by Seb Towers and Ola Kalisz. It is licensed under the terms of the MIT license.

Credits

antibody_shapers was created with cookiecutter and the py-pkgs-cookiecutter template.