🧬 BAPred

Protein-Ligand Binding Affinity Prediction using Graph Neural Networks

High-performance protein-ligand binding affinity prediction model - 2nd place in CASP16 ligand affinity challenge

🌟 Features

• 🏆 CASP16: 2nd place in the prestigious CASP16 ligand affinity prediction challenge
• 🎯 High Accuracy: Graph Neural Network-based architecture for precise binding affinity prediction
• 🔬 Research Ready: Pre-trained models ready for immediate use
• 🛠️ Easy Integration: Simple Python API and command-line interface
• 📈 Scalable: Batch processing for high-throughput screening

🚀 Quick Start

Installation

Choose your preferred installation method:

📦 Option 1: Install from PyPI (Simplest)

pip install bapred

🐍 Option 2: Using Conda (Recommended for Development)

git clone https://github.com/eightmm/BAPred.git
cd BAPred
conda env create -f env.yaml
conda activate BAPred
pip install -e .

🔧 Option 3: From Source

git clone https://github.com/eightmm/BAPred.git
cd BAPred
pip install -r requirements.txt
pip install -e .

🏃‍♂️ Run Your First Prediction

python run_inference.py -r example/1KLT.pdb -l example/ligands.sdf -o results.csv

That's it! 🎉 Your binding affinity predictions will be saved in results.csv.

📋 Usage Examples

Basic Usage

# Predict binding affinities
python run_inference.py -r example/1KLT.pdb -l example/ligands.sdf -o results.csv

Advanced Options

# Use CPU instead of GPU
python run_inference.py -r protein.pdb -l ligands.sdf -o results.csv --device cpu

# Custom batch size for memory optimization
python run_inference.py -r protein.pdb -l ligands.sdf -o results.csv --batch_size 64

# Limit CPU workers for data loading
python run_inference.py -r protein.pdb -l ligands.sdf -o results.csv --ncpu 8

# Specify custom model path
python run_inference.py -r protein.pdb -l ligands.sdf -o results.csv --model_path /path/to/model

Python API

from bapred.inference import inference

# Run prediction programmatically
inference(
    protein_pdb="example/1KLT.pdb",
    ligand_file="example/ligands.sdf",
    output="results.csv",
    batch_size=128,
    ncpu=4,
    model_path="bapred/weight",
    device="cuda"
)

📁 Project Structure

BAPred/
├── 📦 bapred/                 # Main package
│   ├── 🧪 data/               # Data processing modules
│   │   ├── atom_feature.py    # Atomic feature extraction
│   │   ├── data.py           # Dataset handling
│   │   └── utils.py          # Utility functions
│   ├── 🧠 model/              # Neural network models
│   │   ├── GatedGCNLSPE.py   # Gated Graph Convolution
│   │   ├── GraphGPS.py       # Graph GPS architecture
│   │   ├── MHA.py            # Multi-Head Attention
│   │   └── model.py          # Main model wrapper
│   ├── ⚖️ weight/             # Pre-trained weights
│   │   └── BAPred.pth        # Model checkpoint
│   └── 🔮 inference.py       # Inference engine
├── 📝 example/               # Example files
│   ├── 1KLT.pdb             # Sample protein structure
│   └── ligands.sdf          # Sample ligand library
├── 🚀 run_inference.py      # Easy-to-use script
├── 📋 requirements.txt      # Python dependencies
├── 🐍 env.yaml             # Conda environment
└── 📖 README.md            # You are here!

🎯 Model Architecture

BAPred leverages cutting-edge graph neural network architectures:

• 🔗 Graph Convolution: Gated GCN with Laplacian Positional Encoding
• 🌐 Graph GPS: Global attention mechanism for long-range interactions
• 🎭 Multi-Head Attention: Enhanced feature representation
• 🔄 Complex Interactions: Protein-ligand interaction modeling

📊 Input/Output Formats

Input

• Protein: PDB format (.pdb)
• Ligands: SDF (.sdf), MOL2 (.mol2), or text file with paths (.txt)

Output

• CSV/TSV file with columns:
  • Name: Ligand identifier
  • pKd: Predicted binding affinity (pKd scale)
  • Kcal/mol: Binding energy in kcal/mol

🛠️ System Requirements

• Python: 3.11 or higher
• Memory: 4GB RAM minimum (8GB+ recommended)
• GPU: CUDA-compatible GPU (optional, but recommended for speed)
• Storage: 2GB free space

📊 Performance

Dataset	Ligands	Processing Time	Performance
CASP16	Challenge dataset	Competition	🥈 2nd Place
Example	500	~3 minutes	High precision
Custom	Variable	Scales linearly	Research-grade

🏆 CASP16 Achievement

BAPred achieved 2nd place in the CASP16 (Critical Assessment of protein Structure Prediction) ligand affinity prediction challenge, demonstrating its state-of-the-art performance in real-world protein-ligand binding affinity prediction tasks.

🤝 Contributing

We welcome contributions! Please feel free to:

• 🐛 Report bugs
• 💡 Suggest features
• 📖 Improve documentation
• 🔧 Submit pull requests

📄 License

This project is licensed under the Apache License 2.0. See the LICENSE file for details.

📚 Citation

If you use BAPred in your research, please cite:

@software{bapred2024,
  title={BAPred: Protein-Ligand Binding Affinity Prediction using Graph Neural Networks},
  author={Jaemin Sim},
  year={2024},
  url={https://github.com/eightmm/BAPred}
}

🙋‍♀️ Support

• 📖 Documentation: Check this README and code comments
• 🐛 Issues: GitHub Issues
• 💬 Discussions: GitHub Discussions

---

Made with ❤️ for the scientific community

⭐ Star us on GitHub if this project helped you!