BackgroundThe bead-based epitope assay (BBEA) has been used to identify epitope-specific (es) antibodies and successfully utilized to diagnose clinical allergy to milk, egg and peanut. ObjectiveThis study aimed to identify es-IgE, es-IgG4 and es-IgG1 of wheat proteins and determine the optimal peptides to differentiate wheat-allergic from wheat-tolerant using the BBEA. MethodsChildren and adolescents who underwent an oral food challenge to confirm their wheat allergy status were enrolled. Seventy-nine peptides from alpha/beta-gliadin, gamma-gliadin (γ-gliadin), omega-5-gliadin (ω-5-gliadin), high and low molecular weight glutenin were commercially synthesized and coupled to LumAvidin beads. Machine learning (ML) methods were used to identify diagnostic epitopes and performance was evaluated using DeLong’s test. ResultsThe analysis includes 122 children (83 wheat-allergic and 39 wheat-tolerant, 57.4% male). ML coupled with simulations identified wheat es-IgE, but not es-IgG4 or es-IgG1 to be most informative for diagnosing wheat allergy. Higher es-IgE binding intensity correlated with the severity of allergy phenotypes, with wheat anaphylaxis exhibiting the highest es-IgE binding intensity. In contrast, wheat-dependent exercise-induced anaphylaxis (WDEIA) showed lower es-IgG1 binding than all other groups. A set of 4 informative epitopes from ω-5-gliadin, and γ-gliadin were the best predictors of wheat allergy with an AUC of 0.908 (sensitivity=83.4%, specificity=88.4%), higher than the performance exhibited by wheat-specific IgE (AUC=0.646, p < 0.001). The predictive ability of our model was confirmed in an external cohort of 71 patients (29 allergic, 42 non-allergic), with an AUC of 0.908 (sensitivity=75.9%, specificity=90.5%). ConclusionThe wheat BBEA demonstrated greater diagnostic accuracy compared to existing specific IgE tests for wheat allergy.