An immunodominance perspective on a paradoxical phenomenon: discovery and modeling of ragweed and tree sensitization as negative predictors for high mugwort IgE reactivity

Background Mugwort allergy poses a significant disease burden in Northwestern China. While sensitization to mugwort can be confirmed through specific IgE testing, the immunological factors that influence the intensity of the IgE response among sensitized individuals remain poorly understood. This study aimed to explore whether sensitization profiles to other common inhalant allergens are associated with the magnitude of mugwort-specific IgE responses in already-sensitized patients, and to characterize the patterns of co-sensitization that may reflect underlying immune mechanisms. Methods In this retrospective retrospective study, we enrolled 635 mugwort-sIgE-positive patients from a tertiary hospital. Levels of sIgE to multiple inhalant allergens, including ragweed, trees, and dust mites, were measured. The least absolute shrinkage and selection operator (LASSO) regression identified the most predictive features for high mugwort IgE reactivity (grade ≥4). An XGBoost model was developed and validated on a temporal validation set ( n = 159). Model interpretability was achieved using SHapley Additive exPlanations (SHAP), and a clinical nomogram was constructed. Results Three features—ragweed, tree, and dust mite sIgE concentrations—were selected, all exhibiting negative coefficients. The XGBoost model demonstrated excellent discrimination, with an AUC of 0.851 (95% CI: 0.781–0.919) on the test set and 0.827 on the temporal validation set. SHAP analysis revealed the counterintuitive, negative predictive effects of these allergens on high mugwort reactivity, supporting an immunodominance-based “molecular masking” hypothesis. The derived nomogram and risk stratification system effectively stratified patients, with the high-risk group having an actual positive rate of 85%. Decision curve analysis confirmed the clinical utility of the model. Conclusion From an immunodominance perspective, we discovered and modeled sensitization to ragweed, trees, and dust mites as negative predictors for high mugwort reactivity. The interpretable machine learning model provides a precise tool for early identification and stratified management of high-risk mugwort-allergic patients.