genostack / GeoAB

This is the official repository of paper - GeoAB: Towards Realistic Antibody Design and Reliable Affinity Maturation (ICML2024)

Installation

Install requirements

conda create -n geoab python==3.9
conda activate geoab
pip install -r requirements.txt

Dataset

Please follow the data preparation scripts in DyMean, which leads the raw data set as

- all_data
    - RAbD_H3
        - test_processed
            _metainfo
            part_0.pkl
        - train_processed
            ...
        - valid_processed
            ...
        test.json
        train.json
        valid.json
    - SKEMPI
        ...

The processed data can be downloaded from google drive https://drive.google.com/drive/folders/1pNsoGt0gDIJR9EMmmp2pItjZDuLZI3gg. After downloading all_data.zip, unzip it and the processed datasets can be obatained.

Train and evaluate GeoAB-Designer

Run the following command for training:

# Train GeoAB-refiner
python train_refine.py

# Train GeoAB-Initializer
python train_init.py
# After GeoAB-Initializer is trained, train GeoAB-Designer
python train_design.py

For evaluation, run the following command:

# Evaluate GeoAB-Refiner
python eval.py --eval_dir H3_refine --run 1

# Evaluate GeoAB-Designer
python eval.py --eval_dir H3_design

We give the pretrained model in cdrh3.zip, which can be downloaded from https://drive.google.com/drive/folders/1pNsoGt0gDIJR9EMmmp2pItjZDuLZI3gg. You can evaluate the results directly using our pretrained models.

DDG Prediction

For DDG prediction part, our model will be updated through a platform, which will be online soon.

Citation

Please cite the paper if the repository or the paper is helpful to you, as following

TODO