IL-6-Inducing Peptide Prediction Based on 3D Structure and Graph Neural Network

Area of Science:

• Computational biology
• Immunology
• Biochemistry

Background:

• Interleukin-6 (IL-6) is a key glycoprotein in immunity and metabolism, with its expression linked to disease severity.
• IL-6-inducing peptides are vital for developing immunotherapies and diagnostic biomarkers.
• Current prediction methods lack 3D structural information and rely on prior knowledge.

Purpose of the Study:

• To propose a novel method, DGIL-6, for predicting IL-6-inducing peptides.
• To integrate 3D structural information and graph neural networks for enhanced prediction accuracy.
• To address limitations of existing machine learning approaches.

Main Methods:

• Representing peptide sequences as graphs with amino acids as nodes.
• Utilizing predicted residue contact graphs for adjacency matrices.
• Employing ESM-1b for amino acid feature extraction.
• Applying a dual-channel Graph Attention Network (GAT) and Graph Convolutional Network (GCN) approach.

Main Results:

• DGIL-6 effectively integrates 3D structural data and graph neural networks.
• The dual-channel GAT-GCN method captures node weights and information updates.
• Experimental validation, including cross-validation and independent testing, confirms the method's efficacy.

Conclusions:

• DGIL-6 offers a significant advancement in predicting IL-6-inducing peptides.
• The integration of 3D structural information is crucial for functional peptide prediction.
• This approach holds promise for future immunotherapy and biomarker development.