Generative model for monoclonal antibodies

This repo contains code for the paper [Energy-Based generative models for monoclonal antibodies]. Some of the code was adapted from the repository [GflowNet for Biological Sequence Design] (https://github.com/MJ10/BioSeq-GFN-AL) See Licence for more information

Setup

The code has been tested with Python 3.7 with CUDA 11.3 and CUDNN 8.0.

1. We recommand setting up an anaconda environment before running the code
2. Before installing the requirements, ensure you have a c++ compiler available on your machine (apt-get install build-essential on ubuntu)
3. Install the dependencies (pip install -r requirements.txt)
4. Install anarci (conda install bioconda::anarci)
5. Run the script download_data_and_embeddings to download the necessary ESM embeddings of AAYL49 and the sequences generated by our sampling methods

Running the code

mcmc_covid.py, mcmc_true_aff.py, and mcmc_true_aff_hard.py are the entry points for the generation of sequences using MCMC. run_covid.py, run_true_aff.py, and run_true_aff_hard.py are the entry points for the generation of sequences using GFlowNet. antBO_simple.py, antBO_hard.py are the entry points for the generation of sequences using antBO.

Please reach out to Paul Pereira, [email protected] for any issues, comments, questions or suggestions.