BAPred: Protein-ligand Binding Affinity Prediction

This repository contains code for predicting the binding affinity between protein-ligand pairs using GNN (Graph Neural Network) models. The model predicts the binding affinity score for each ligand pose relative to the receptor.

Installation

1. Clone the repository:

git clone https://github.com/eightmm/BAPred.git
cd BAPred

2. Set up a Python environment and install dependencies: You can use the provided env.yaml to create a conda environment:

conda env create -f env.yaml
conda activate BAPred

Usage

This repository provides a script to predict the binding affinity of protein-ligand pairs using a pretrained model.

Running the Inference

To run the inference and predict binding affinity, use the following command:

python inference.py -r ./example/1KLT.pdb -l ./example/ligands.sdf -o ./result.csv --model_path ./weight --device cuda

Where:

• -r specifies the receptor protein PDB file.
• -l specifies the ligand SDF file.
• -o specifies the output CSV file for results.
• --model_path specifies the directory containing the model weight (BAPred.pth).
• --device specifies whether to use cuda or cpu.

Output

The output will be saved in the specified CSV file and will contain the following columns:

• Name: Name or index of the ligand.
• Affinity: Predicted binding affinity score for the ligand pose.

File Structure

.
├── data
│   ├── atom_feature.py        # Atom feature extraction
│   ├── data.py                # Data loading and preprocessing
│   └── utils.py               # Utility functions
├── example
│   ├── 1KLT.pdb               # Example receptor PDB file
│   ├── ligands.sdf            # Example ligand SDF file
│   └── run.sh                 # Example script to run inference
├── inference.py               # Inference script for binding affinity prediction
├── LICENSE                    # License file
├── model
│   ├── GatedGCNLSPE.py        # GNN model implementation
│   ├── GraphGPS.py            # GraphGPS model architecture
│   ├── MHA.py                 # Multi-Head Attention module
│   ├── model.py               # Model wrapper class
├── README.md                  # This README file
└── weight
    └── BAPred.pth             # Saved weight for the binding affinity prediction model

Example

Below is an example of how to run the code:

python inference.py     -r ./example/1KLT.pdb     -l ./example/ligands.sdf     -o ./result.csv     --batch_size 128     --model_path ./weight     --device cuda

The example receptor 1KLT.pdb and ligand ligands.sdf are provided in the example/ directory. This command will generate a CSV file named result.csv containing the predicted binding affinity values for each ligand pose.

Models

The prediction model is based on Graph Neural Networks (GNNs) with additional layers such as GraphGPS and Multi-Head Attention. The model takes the protein and ligand graphs as input and outputs the predicted binding affinity for each ligand pose.

• Binding Affinity Model (BAPred.pth): Predicts the binding affinity of the ligand pose.

The model architectures are defined in model/GatedGCNLSPE.py, model/GraphGPS.py, and model/MHA.py.