Identification of Protein-Nucleotide Binding Residues With Deep Multi-Task and Multi-Scale Learning

Area of Science:

• Computational biology
• Bioinformatics
• Molecular biology

Background:

• Accurate identification of protein-nucleotide binding residues is crucial for understanding protein function and enabling drug discovery.
• Current computational methods face challenges in extracting discriminative features and integrating diverse data from nucleotide binding sites.

Purpose of the Study:

• To develop a novel computational predictor, NucMoMTL, for accurate identification of protein-nucleotide binding residues.
• To address limitations in feature extraction and data assimilation in existing prediction methodologies.

Main Methods:

• NucMoMTL employs a pre-trained language model for robust protein sequence embedding.
• It utilizes deep multi-task and multi-scale learning with parameter-based orthogonal constraints.
• The method integrates auxiliary information from various nucleotide binding residues to extract shared representations.

Main Results:

• NucMoMTL demonstrated superior performance compared to state-of-the-art methods on benchmark datasets.
• The predictor achieved an average Area Under the Receiver Operating Characteristic curve (AUROC) of 0.961.
• An average Area Under the Precision-Recall Curve (AUPRC) of 0.566 was obtained.

Conclusions:

• NucMoMTL serves as a reliable computational tool for identifying protein-nucleotide binding residues.
• The tool has the potential to significantly facilitate protein functional annotation and drug discovery efforts.
• The source code and dataset are publicly available for further research and application.