ProtLGN

Protein Engineering with Lightweight Graph Denoising Neural Networks

Explore the docs »

View Demo · Report Bug · Request Feature

About The Project

ProtLGN is pre-trained on wild-type proteins for AA-type denoising tasks with equivariant graph neural networks to derive the joint distribution of the recovered AA types (red).

For a protein to mutate, the predicted probabilities suggest the fitness score for associated mutations (blue).

With additional mutation evaluations from wet biochemical assessments, the pre-trained model can be updated to better fit the specific protein and protein functionality (green).

(back to top)

📄 News

• [2024.06.06] We recently developed two more advanced protein engineering tools named ProtSSN and ProSST for zero-shot prediction. We recommend you try the new models!

Getting Started

Please follow these simple example steps to get start! 😊

Conda Enviroment

Please make sure you have installed Anaconda3 or Miniconda3.

Enviroment.

conda env create -f environment.yaml
conda activate protlgn
pip install torch_scatter torch_sparse torch_cluster -f https://data.pyg.org/whl/torch-2.3.0+cu121.html

Pre-train ProtLGN

Step 1: get raw dataset

We use the dataset from CATH 4.2, you can download from https://www.cathdb.info/.

mkdir -p data/cath_k10/raw
cd data/cath_k10/raw
wget https://huggingface.co/datasets/tyang816/cath/blob/main/dompdb.tar
# or wget https://lianglab.sjtu.edu.cn/files/ProtSSN-2024/dompdb.tar
tar -xvf dompdb.tar

Step 2: build graph dataset

see script/build_cath_dataset.sh

Step 3: run pre-train

see run_pretrain.sh

Zero-shot prediction for mutant sequences

You can use your own checkpoint for zero-shot inference.

Step 1: Prepare mutant dataset

Data map:

|—— eval_dataset
|——|—— DATASET
|——|——|—— Protein1
|——|——|——|—— Protein1.tsv (DMS file)
|——|——|——|—— Protein1.pdb (pdb file)
|——|——|——|—— Protein1.fasta (sequence)
|——|——|—— Protein2
|——|——|——|...

see script/build_mutant_dataset.sh

Step 2: Zero-shot

see script/mutant_predict.sh

CUDA_VISIBLE_DEVICES=0 python mutant_predict.py \
    --checkpoint ckpt/ProtLGN.pt \
    --c_alpha_max_neighbors 10 \
    --gnn egnn \
    --use_sasa \
    --layer_num 6 \
    --gnn_config src/Egnnconfig/egnn_mutant.yaml \
    --mutant_dataset data/example

Contributing

Please cite our paper:

@article{zhou2024protlgn,
  title={Protein engineering with lightweight graph denoising neural networks},
  author={Zhou, Bingxin and Zheng, Lirong and Wu, Banghao and Tan, Yang and Lv, Outongyi and Yi, Kai and Fan, Guisheng and Hong, Liang},
  journal={Journal of Chemical Information and Modeling},
  volume={64},
  number={9},
  pages={3650--3661},
  year={2024},
  publisher={ACS Publications}
}

@article{tan2023protssn
  title={Semantical and Topological Protein Encoding Toward Enhanced Bioactivity and Thermostability},
  author={Tan, Yang and Zhou, Bingxin and Zheng, Lirong and Fan, Guisheng and Hong, Liang},
  journal={bioRxiv},
  pages={2023--12},
  year={2023},
  publisher={Cold Spring Harbor Laboratory}
}

(back to top)

License

Distributed under the MIT License. See LICENSE.txt for more information.

(back to top)