HLAIImaster: a deep learning method with adaptive domain knowledge predicts HLA II neoepitope immunogenic responses

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Accurate prediction of neoepitopes that stimulate CD4+ T cell responses is crucial for developing effective cancer immunotherapies and vaccines.
• Current methods for predicting neoepitopes presented by human leukocyte antigen (HLA) class II molecules are limited by data scarcity and algorithmic constraints.
• Identifying HLA class II-restricted peptides is essential for understanding T cell-mediated immunity in cancer and for designing personalized treatments.

Purpose of the Study:

• To develop a robust computational framework for predicting exogenous HLA class II-restricted peptides across a broad human population.
• To enhance the accuracy of neoepitope immunogenicity prediction by integrating mass spectrometry data, peptide processing, and gene expression.
• To introduce HLAIImaster, a novel deep learning model designed to improve the prediction of neoepitope presentation by HLA class II molecules.

Main Methods:

• Utilized mass spectrometry data to profile over 223,000 eluted ligands across HLA-DR, -DQ, and -DP alleles.
• Developed HLAIImaster, an attention-based deep learning framework incorporating adaptive domain knowledge.
• Integrated peptide processing information and gene expression data with eluted ligand profiles to train the deep learning model.

Main Results:

• HLAIImaster demonstrated significant improvements in positive predictive value compared to existing tools in neoantigen studies.
• The framework accurately identifies neoepitope immunogenicity by leveraging diverse biological characteristics and advanced deep learning.
• The study successfully exploited the immunogenic neoepitope repertoire of cancers, validated by mass spectrometry data.

Conclusions:

• HLAIImaster effectively bridges the gap between neoantigen biology and clinical applications by accurately predicting neoepitope immunogenicity.
• The developed framework facilitates the creation of "just-in-time" personalized cancer vaccines.
• This advancement holds promise for future neoantigen-based therapies, potentially offering greater clinical benefit to patients.