DSSA-PPI: enhancing binding affinity change prediction upon protein mutations using disentangled structure-sequence aware attention

This repository closely reproduce the experiments of: Juhong Wu et. al. DSSA-PPI: enhancing binding affinity change prediction upon protein mutations using disentangled structure-sequence aware attention. Now it have accepted by Chemical Science! DSSA-PPI

Getting started with this repo

Main requirements

• Pytorch(v2.0.1). Pytorch with GPU version. Use to model, train, and evaluate the actual neural networks.
• Graphein. Use to process 3D structures of protein complex.
• Equiformer-pytorch. The backbone of structure encoder.
• ESM. For protein sequence embedding.

To install DSSA-PPI

git clone https://github.com/Agitw/DSSA-PPI
cd DSSA-PPI

# Ensure Conda is installed before running the next commands
# If Conda is not installed, download it from: https://docs.conda.io/en/latest/miniconda.html
conda env create -f environment.yml
conda activate DSSA-PPI

Download data.

The precomputed data using to reproduce the results of our paper is sourced at DSSA-PPI data. This repository includes the pretrained model weights for DSSA-PPI and preprocessed model inputs data.

wget https://zenodo.org/records/15088340/files/data.tar.gz
tar -zxvf data.tar.gz # need to place in DSSA-PPi/data

Training and evaluation.

Training on benchmark datasets S1131, S4169 or M1707

cd run_scripts
bash run.sh <CUDA_id> <task [S1131, S4169, M1707]> 

Validation on benchmark datasets S1131, S4169 or M1707

cd validation
python validate.py --kind S1131  # Default is S1131

Evaluation on RBM of SARS-CoV-2 interacting with ACE2

cd validate_on_RBM
python rbm_scanning.py

License

DSSA-PPI is released under an MIT License. DSSA-PPI: enhancing binding affinity change prediction upon protein mutations using disentangled structure-sequence aware attention.