Relation-Aware Equivariant Graph Networks for Epitope-Unknown Antibody Design and Specificity Optimization
Lirong Wu, Haitao Lin, Yufei Huang, Zhangyang Gao, Cheng Tan, Yunfan Liu, Tailin Wu, Stan Z. Li. In AAAI, 2025.
The script for environment setup is available in scripts/setup.sh
, please install the dependencies before running code.
bash scripts/setup.sh
Dataset | Download Script |
---|---|
SAbDab | scripts/prepare_data_kfold.sh |
RAbD | scripts/prepare_data_rabd.sh |
SKEMPI v2 | scripts/prepare_data_skempi.sh |
We have provided the summary data used in our paper from SAbDab, RAbD, SKEMPI_V2 in the summaries folder, and you can use the above scripts to download the required data. The processed data can be downloaded from Google Drive. After downloading data for RAAD.zip
, unzip it and replace the summaries folder with the processed datasets.
python -B train.py --cdr_type 1 --optimization 0
where cdr_type
denotes the type of CDR on the heavy chain.
The customized hyperparameters can be searched in the space provided by . /configs/search_space.json
.
python -B train.py --optimization 1
The customized hyperparameters can be searched in the space provided by . /configs/search_space.json
.
python -B train.py --optimization 2
The customized hyperparameters can be searched in the space provided by . /configs/search_space.json
.
If you are interested in our repository and our paper, please cite the following paper:
@article {lin2024geoab,
author = {Lin, Haitao and Wu, Lirong and Huang, Yufei and Liu, Yunfan and Zhang, Odin and Zhou, Yuanqing and Sun, Rui and Li, Stan Z.},
title = {GeoAB: Towards Realistic Antibody Design and Reliable Affinity Maturation},
year = {2024},
booktitle={International Conference on Machine Learning},
URL = {https://www.biorxiv.org/content/early/2024/05/17/2024.05.15.594274},
eprint = {https://www.biorxiv.org/content/early/2024/05/17/2024.05.15.594274.full.pdf}
}
If you have any issue about this work, please feel free to contact me by email:
- Lirong Wu: wulirong@westlake.edu.cn
Many thanks for the code provided below