μProtein is a general framework designed to accelerate protein engineering by integrating μFormer, a deep learning model for accurate mutational effect prediction, with μSearch, a reinforcement learning algorithm tailored for efficient navigation of the protein fitness landscape.
For more details, please refer to our paper in Nature Machine Intelligence.
This repository contains the following components:
pmlm/– Protein language model pretrainingmu-former/– Fitness landscape modeling using the pretrained protein language modelmu-search/– Navigating the constructed fitness landscape oraclepretrained/– Pretrained PMLM model checkpoint (stored using Git LFS).
For more details, refer to the respective README files:
If you are using our code or model, please cite the following paper:
@article{sun2025accelerating,
title={Accelerating protein engineering with fitness landscape modelling and reinforcement learning},
author={Sun, Haoran and He, Liang and Deng, Pan and Liu, Guoqing and Zhao, Zhiyu and Jiang, Yuliang and Cao, Chuan and Ju, Fusong and Wu, Lijun and Liu, Haiguang and others},
journal={Nature Machine Intelligence},
pages={1--15},
year={2025},
publisher={Nature Publishing Group UK London}
}This repository is licensed under the MIT License.
For questions or collaborations, please contact the authors via email or open an issue in this repository.