:loudspeaker: PocketGen: Generating Full-Atom Ligand-Binding Protein Pockets

Environment

Install conda environment via conda yaml file

conda env create -f pocketgen.yaml
conda activate pocketgen

Install via Conda and Pip

conda create -n targetdiff python=3.8
conda activate targetdiff
conda install pytorch pytorch-cuda=11.6 -c pytorch -c nvidia
conda install pyg -c pyg
conda install rdkit openbabel tensorboard pyyaml easydict python-lmdb -c conda-forge
conda install -c conda-forge openmm pdbfixer flask
conda install -c conda-forge numpy swig boost-cpp sphinx sphinx_rtd_theme
pip install meeko==0.1.dev3 wandb scipy pdb2pqr vina==1.2.2 
python -m pip install git+https://github.com/Valdes-Tresanco-MS/AutoDockTools_py3

Benchmark Datasets

We use CrossDocked and Binding MOAD datasets to benchmark pocket generation.

CrossDocked

We download and process the CrossDocked dataset as described by the authors of TargetDiff
Firstly download the crossdocked_v1.1_rmsd1.0.tar.gz and split_by_name.pt and put it under the ./data directory.
Use the following commands to extract pockets, create index_seq.pkl, and split the dataset.

python data_preparation/extract_pockets.py
python data_preparation/split_pl_dataset.py

Binding MOAD

We download and process the Binding MOAD dataset following the authors of DiffSBDD Download the dataset

wget http://www.bindingmoad.org/files/biou/every_part_a.zip
wget http://www.bindingmoad.org/files/biou/every_part_b.zip
wget http://www.bindingmoad.org/files/csv/every.csv

unzip every_part_a.zip
unzip every_part_b.zip

Process the raw data using

python -W ignore process_bindingmoad.py <bindingmoad_dir>

Use the following commands to extract pockets, create index_seq.pkl, and split the dataset.

python data_preparation/extract_pockets_moad.py
python data_preparation/split_pl_dataset_moad.py

Processed datasets

We also provide the processed datasets for training from scratch at zenodo

For each dataset, it requires the preprocessed .lmdb file and split file _split.pt

Benchmark Results

Benchmarking PocketGen and other approaches for pocket generation on two datasets. Reported are average and standard deviation values across three independent runs. The best results are bolded.

Training

Train on CrossDocked:

python train_recycle.py --config ./config/train_model.yml

Train on Binding MOAD:

python train_recycle.py --config ./config/train_model_moad.yml

Model Checkpoints

Pretrained checkpoint on the CrossDocked training dataset: checkpoint.pt

Generation

python generate_new.py

We provide one example of the generated pocket for pdbid-2p16 and visualize the interactions with plip

Self-Consistency Evaluation

The code to compute scRMSD, scTM, and pLDDT can be found at eval.

Acknowledgement

This project draws in part from TargetDiff and ByProt, supported by MIT License and Apache-2.0 License. Thanks for their great work and code!

Contact

Zaixi Zhnag ([email protected])

Sincerely appreciate your suggestions on our work!

License

This project is licensed under the terms of the MIT license. See LICENSE for additional details.

Reference

@article{zhang2024pocketgen,
  title={PocketGen: Generating Full-Atom Ligand-Binding Protein Pockets},
  author={Zhang, Zaixi and Shen, Wanxiang and Liu, Qi and Zitnik, Marinka},
  journal={arXiv},
  url={},
  year={2024}
}