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

## TL;DR

This paper introduces DSCA-HLAII, a new AI model that accurately predicts how peptides bind to HLA class II molecules, which is important for understanding immune responses and drug development.

## Contribution

The novel dual-stream cross-attention mechanism integrates semantic and sequence features for improved peptide-HLA-II interaction prediction.

## Key findings

- DSCA-HLAII outperforms existing methods in predicting peptide-HLA-II interactions and presentation.
- The model accurately identifies key binding sites and supports immunogenicity risk assessment of antibodies.
- A public web server is provided for practical application of the model.

## Abstract

The interaction between peptides and human leukocyte antigen class II (HLA-II) molecules plays a pivotal role in adaptive immune responses, as HLA-II mediates the recognition of exogenous antigens and initiates T cell activation through peptide presentation. Accurate prediction of peptide-HLA-II binding serves as a cornerstone for deciphering cellular immune responses, and is essential for guiding the optimization of antibody therapeutics. Researchers have developed several computational approaches to identify peptide-HLA-II interaction and presentation. However, most computational approaches exhibit inconsistent predictive performance, poor generalization ability and limited biological interpretability.

In this study, we present DSCA-HLAII, a novel predictive framework for peptide-HLA-II interactions and presentation based on a dual-stream cross-attention architecture. The framework proposes a dual-stream cross-attention (DSCA) mechanism to integrate pre-trained semantic embedding ESMC with sequence-level ONE-HOT features. The DSCA mechanism effectively models the interaction dynamics between peptides and HLA-II molecules, enabling the precise identification of key binding sites. Experimental results demonstrate that DSCA-HLAII consistently surpasses existing state-of-the-art approaches, demonstrating high accuracy and robustness in predicting peptide-HLA-II interactions and presentation. We further demonstrate the capability of DSCA-HLAII for predicting peptide binding cores and assessing antibody immunogenicity, which is expected to advance artificial intelligence-based peptide drug discovery.

This paper proposes a novel predictive framework for peptide-HLA-II interactions and presentation based on a dual-stream cross-attention architecture (DSCA-HLAII). DSCA-HLAII is a unified predictive framework that integrates sequence-based ONE-HOT features with pre-trained semantic embeddings (ESMC) to construct a comprehensive hybrid representation of peptides and full-length HLA-II sequences. By introducing a Dual-Stream Cross-Attention (DSCA) module, the model enables fine-grained characterization of peptide–HLA-II interactions and assigns differential scores across sequence positions, thereby improving the identification of critical binding sites and enhancing generalization. DSCA-HLAII simultaneously predicts peptide presentation probability and binding core location, and further supports systematic assessment of antibody immunogenicity risk. Extensive experiments on multiple warm-start test datasets and cold-start test datasets demonstrate that DSCA-HLAII surpasses existing state-of-the-art methods in both accuracy and robustness. Additionally, a publicly accessible web server (http://bliulab.net/DSCA-HLAII) has been established to facilitate practical application.