DSCA-HLAII: A dual-stream cross-attention model for predicting peptide-HLA class II interaction and presentation.

Area of Science:

• Immunoinformatics
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Peptide-HLA-II interactions are crucial for adaptive immunity and T cell activation.
• Accurate prediction of these interactions is vital for understanding immune responses and developing antibody therapeutics.
• Current computational methods for predicting peptide-HLA-II binding often lack consistent performance and interpretability.

Purpose of the Study:

• To develop a novel predictive framework, DSCA-HLAII, for peptide-HLA-II interactions and presentation.
• To improve the accuracy, generalization ability, and biological interpretability of computational approaches in this field.
• To advance AI-driven peptide drug discovery.

Main Methods:

• Development of DSCA-HLAII, a framework utilizing a dual-stream cross-attention architecture.
• Integration of pre-trained semantic embedding (ESMC) with sequence-level ONE-HOT features via a dual-stream cross-attention (DSCA) mechanism.
• Modeling of interaction dynamics between peptides and HLA-II molecules to identify key binding sites.

Main Results:

• DSCA-HLAII demonstrates superior performance compared to state-of-the-art methods in predicting peptide-HLA-II interactions and presentation.
• The framework exhibits high accuracy and robustness.
• Successfully predicted peptide binding cores and assessed antibody immunogenicity.

Conclusions:

• DSCA-HLAII offers a significant advancement in predicting peptide-HLA-II interactions.
• The framework's capabilities are expected to accelerate AI-based peptide drug discovery.
• Provides a more accurate and interpretable tool for immunoinformatics research.