Association of allergen signatures with individualized allergic phenotypes.

Abstract

BACKGROUND: Allergen sensitization patterns are heterogeneous, and their clinical relevance is often obscured by extensive cross-reactivity. We applied non-negative matrix factorization (NMF) to disentangle overlapping immunoglobulin E (IgE) signals and define clinically meaningful allergen signatures in a large Korean cohort. METHODS: We analyzed 45,065 patients who underwent multiplex allergen testing (35 inhalants and food components) between 2010 and 2025. Class-scaled specific IgE values (0-6) were factorized by NMF (k = 4). Signature weights were related to asthma, allergic rhinitis, and atopic dermatitis using multivariable logistic regression and to peripheral eosinophil counts and total IgE using age- and sex-adjusted linear models. RESULTS: Four signatures-mite, grass/weed, pet, and tree-explained 77.7 % of the variance in sensitization. The mite signature predominated (57.6 % of patients) and was strongly associated with allergic rhinitis (adjusted OR: 7.21, 95 % CI: 5.66-9.16), as well as marked increases in eosinophils and total IgE. The pet signature was the strongest predictor of asthma (OR: 8.90, 6.48-12.24). The tree signature showed the strongest association with atopic dermatitis (OR: 6.27, 3.81-10.32) and broader multisystem allergic morbidity. The grass/weed signature exhibited a biphasic age trajectory with a late-adult resurgence but had modest clinical impact. All signatures were significant and graded as determinants of blood eosinophil counts and IgE levels. CONCLUSIONS: Data-driven factorization of multiplex IgE panels yields portable allergen signatures that refine attribution of asthma, allergic rhinitis, and atopic dermatitis and link serologic patterns to systemic inflammation.