This repository contains code for the paper "De novo antibody design with SE(3) diffusion". It is adapted from the original FrameDiff code.
To download the model weights and dataset visit the HuggingFace repo, or run
git clone https://huggingface.co/AmelieSchreiber/IgDiffThen move the weights into the weights folder, and the samples into the igdiff_samples folder.
To install the package, run
pip install -e .
If needed, you can create a conda environment with all the required dependencies using
conda env create -f se3.yml
You can generate unconditioned antibodies with
python experiments/inference_se3_diffusion.py
which will create a folder in the inference directory with generated structures.