In Silico Identification of HLA-DRB5*01:01-Restricted Peanut Peptides

Peanut allergy is a major food allergy associated with severe IgE-mediated hypersensitivity reactions and strong T-cell responses to peanut-derived epitopes. In this study, complementary computational approaches were developed to identify digestion-stable peanut peptides binding to HLA-DRB5*01:01, an HLA class II molecule associated with peanut allergy. A dataset of experimentally validated binders and non-binders was used to construct one sequence logo model and three machine learning (ML) models based on Random Forest (RF), Support Vector Machine (SVM), and Gradient Boosting (GB) algorithms. Peptide sequences were encoded using Wold’s z-scale descriptors representing hydrophobic, steric, and electronic amino acid properties. All ML models showed excellent predictive performance, with ROC AUC values between 0.978 and 0.980, while the SVM model achieved the best overall accuracy (95.8%). Permutation-importance analysis identified the dominant peptide properties for HLA-DRB5 binding. The developed models were applied to major peanut allergens Ara h 1, Ara h 2, Ara h 3, and Ara h 6 following simulated gastrointestinal digestion. Multiple consensus HLA-DRB5-binding peptides were identified, including several experimentally reported allergenic epitopes. The identified peptides should be regarded as candidate HLA-DRB5*01:01-restricted CD4+ T-cell epitopes that warrant further experimental validation using peptide–HLA binding assays and T-cell functional studies.