Proteins are essential biomolecules that regulate virtually all biological processes. Determining their function based solely on their sequence remains a significant challenge in bioinformatics. Recent advances have led to the development of models like ESM-2, which learn protein representations to assist in various bioinformatics tasks. However, most current methods focus primarily on sequence information, often neglecting the structural characteristics of proteins.
We introduce ProGraphTrans, a novel approach that combines Graph Neural Networks (GNNs) with Transformers to enhance protein representation learning by integrating both sequence and structural features. Our method leverages ESM-2 embeddings and amino acid contact profiles to build richer representations, improving the performance of protein-related tasks.
- Python 3.7+
- PyTorch 1.10+
- ESM-2
- transformerCPI
To get started with ProGraphTrans, clone the repository and install the required dependencies.
# Clone the repository
git clone https://github.com/lftxd1/ProGraphTrans.git
cd ProGraphTrans
# Install dependencies
pip install -r requirements.txtProGraphTrans has been validated on multiple protein-related tasks:
- DNA-Binding Protein Prediction: Predict whether a protein binds to DNA based on its sequence and structural features.
- Protein-Protein Interaction Prediction: Predict the interaction between two proteins.
- Protein Solubility Prediction: Predict the solubility of a protein in different conditions.
- Compound-Protein Interaction Prediction: Predict the interaction between a compound and a protein.
We welcome contributions to improve ProGraphTrans. If you want to add new features, fix bugs, or improve the documentation, feel free to submit a pull request.
This project is licensed under the MIT License - see the LICENSE file for details.