Venus-MAXWELL

Source code of Venus-MAXWELL: Efficient Learning of Protein-Mutation Stability Landscapes using Protein Language Models

If you find this work useful, please star the repo! (Click the star button in the top right corner of the page)

Thank you! (づ｡◕‿‿◕｡)づ

Recommended environment

It is highly recommended to start a new conda environment from scratch due to potential CUDA compatability issues between pytorch and the pytorch-geometric package required for the inverse folding model.

To set up a new conda environment with required packages, run the following commands:

conda create -n maxwell python=3.10
conda activate maxwell

# Install pytorch
pip install torch==2.6.0 torchvision==0.21.0 torchaudio==2.6.0 --index-url https://download.pytorch.org/whl/cu118

# Install dependencies
pip install lightning==2.4
pip install numpy==1.26.0
pip install biopython==1.85
pip install pandas==2.1.0

# Install esm-inverse-folding dependencies
pip install biotite==0.40.0
pip install torch_geometric
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.6.0+cu118.html
pip install fair-esm

# Install esm-2 and prosst dependencies (optional)
pip install transformers

Clone the repository

git clone https://github.com/ai4protein/Venus-MAXWELL.git
cd Venus-MAXWELL

Generate ddG predictions of all possible single mutations of a protein from PDB file with Venus-MAXWELL(ESM-IF)

Download the weights from here and put them in the weights folder.

ls weights

should output:

esmif-maxwell.ckpt

python predict_ddg.py --pdb_file example_data/fireprotdb_1AG2_ddG.pdb --ckpt_path weights/esmif-maxwell.ckpt --output_file example_data/fireprotdb_1AG2_ddG.csv --device cpu

Train and Test Venus-MAXWELL (ESM-2) on the ddg_datasets

python check_dataset.py --dataset_path example_datasets/ddg_datasets/train
python check_dataset.py --dataset_path example_datasets/ddg_datasets/valid
python check_dataset.py --dataset_path example_datasets/ddg_datasets/test
python -m maxwell.esm2.train --train_folder example_datasets/train --valid_folder example_datasets/valid --test_folder example_datasets/test

Citation

If you find this work useful, please cite:

@article {Yu2025.05.30.656964,
	author = {Yu, Yuanxi and Jiang, Fan and Ma, XinZhu and Zhang, Liang and Zhong, Bozitao and Ouyang, Wanli and Fan, Guisheng and Yu, Huiqun and Hong, Liang and Li, Mingchen},
	title = {Venus-MAXWELL: Efficient Learning of Protein-Mutation Stability Landscapes using Protein Language Models},
	elocation-id = {2025.05.30.656964},
	year = {2025},
	doi = {10.1101/2025.05.30.656964},
	publisher = {Cold Spring Harbor Laboratory},
	URL = {https://www.biorxiv.org/content/early/2025/06/02/2025.05.30.656964},
	eprint = {https://www.biorxiv.org/content/early/2025/06/02/2025.05.30.656964.full.pdf},
	journal = {bioRxiv}
}