GraphNABP: Identifying nucleic acid-binding proteins with protein graphs and protein language models

Area of Science:

• Computational biology
• Structural bioinformatics
• Drug discovery

Background:

• Nucleic acid-binding proteins (NABP) are crucial for biological processes.
• Existing sequence-based NABP prediction methods often neglect structural information.
• The potential of protein language models (pLM) for NABP prediction remains underexplored.

Purpose of the Study:

• To develop a novel framework, GraphNABP, for predicting NABPs.
• To integrate sequence and predicted 3D structure information for enhanced NABP identification.
• To leverage protein language models and structural data in NABP prediction.

Main Methods:

• Utilized ProtT5 protein language model for sequence embeddings.
• Generated protein molecular graphs from predicted 3D structures (e.g., AlphaFold).
• Employed Graph Attention (GAT) and Bidirectional Long Short-Term Memory (BiLSTM) networks for feature enhancement.

Main Results:

• GraphNABP integrates sequence and predicted 3D structural data.
• The framework outperforms existing state-of-the-art methods on independent test sets.
• Demonstrated the effectiveness of pLM embeddings and structural information for NABP prediction.

Conclusions:

• GraphNABP represents a significant advancement in computational NABP identification.
• The integration of pLMs and predicted structures is highly effective for NABP prediction.
• This approach holds promise for accelerating drug discovery and understanding biological mechanisms.