We propose a novel PPI site prediction framework, ProtFormer-Site, which utilizes large protein language model, an efficient parameter fine-tuning strategy, and the ProtFormer backbone. ProtFormer-Site demonstrated outstanding performance across all evaluation metrics on three benchmark datasets, with Matthews correlation coefficient (MCC) improvements ranging from 22.4% to 61.5% across different datasets. These results demonstrate that ProtFormer offers significant advantages in PPI site prediction, providing a more accurate and efficient solution.
git clone https://github.com/ISYSLAB-HUST/ProtFormer-Site.git
cd ProtFormer-Siteconda env create -f environment.yml
conda activate ProtFormerSiteThe pre-trained models and parameters are placed in the weight folder and the config folder.
We provide test script for users to evaluate the prediction result.
# example: run Single_DeepPPIS test
python predict.py --config ./config/Single_DeepPPISP.yamlIf you encounter any problems, please open an issue.
This project is licensed under the MIT License for the code and a custom license for the parameter files.
The code in this project is licensed under the MIT License. See the LICENSE file for more details.
The parameter files in this project are licensed under a custom license. Educational use is free of charge, while commercial use requires a commercial license. See the PARAMETER_LICENSE file for more details.
ProtFormer-Site with and/or references the following separate libraries and packages:
If you use this code or one of our pretrained models for your publication, please cite our paper:
@article{wang2024,
title={Ultra-fast and Accurate Prediction of Protein-protein Interaction Sites Using Protein Language Models and ProtFormer}
}