Contrastive Machine Learning Reveals the Molecular Profile of Omalizumab Responders in Type 2 Asthma

Résumé fourni par la source

Abstract RATIONALE Asthma, the most common chronic childhood disease, consists of heterogeneous subtypes that respond differently to therapies. Current biomarkers, including blood eosinophil count (BEC) and total immunoglobulin E (IgE) level, inform the selection of type 2 (T2)-targeted treatments (i.e., biologics) but their predictive accuracy varies across diverse populations. Precision medicine efforts are hindered by unrecognized heterogeneity within asthma endotypes and their associated biomarkers. METHODS We developed Phenotype Aware Component Analysis (PACA), a contrastive machine-learning approach, to isolate disease-specific heterogeneity in DNA methylation (DNAm) data from pediatric asthma cohorts. Using whole-blood DNAm from Latino (discovery; n=1,016) and African American (replication; n=756) cohorts, we applied PACA to identify the primary latent axis of variation distinguishing asthma patients from healthy controls. This yielded a DNAm stratification score based on 7,662 CpGs, which we applied to predict bronchodilator response (BDR) to albuterol. We also examined its associations with clinical variables, eosinophil-specific DNAm, and gene expression, and assessed its predictive value for omalizumab response in an independent cohort (Upchurch et al.). RESULTS While BEC and IgE correlate with BDR overall, their predictive value for BDR is observed only in patients with high DNAm scores. BEC correlates with BDR in patients with upper-quartile DNAm scores (odds ratio [OR] for response 1.12; 95% CI [1.04, 1.22]; P=7.9e-4) but not lower-quartile scores (OR 1.05; 95% CI [0.95, 1.17]; P=0.21). Similarly, IgE correlates with BDR in above-median scores (OR 1.42; 95% CI [1.24, 1.63]; P=3.9e-7) but not below-median scores (OR 1.05; 95% CI [0.92, 1.2]; P=0.57). These findings remain consistent within the T2-high endotype but not in T2-low patients, suggesting that our DNAm score identifies previously unrecognized heterogeneity in T2-high asthma. T2-high patients with high DNAm scores display clinical features associated with biologic therapy response, including higher exacerbation scores, lower body mass index (BMI), recent oral corticosteroid use, and reduced lung function. T2-high patients with high DNAm scores exhibited eosinophil-specific hypermethylation, notably in the hallmark eosinophilic inflammation marker CLC, and overexpressed ten genes, including TNFRSF13B and TNFRSF17, which are linked to B cell function and antibody production. Higher pretreatment expression of these genes showed a suggestive association with omalizumab response among T2-high patients (P=0.088; n=40). CONCLUSIONS Our DNAm score enhances the clinical utility of existing biomarkers and captures previously unrecognized heterogeneity within T2 asthma. This approach enables refined patient stratification, improving precision medicine strategies and ensuring broader applicability across diverse populations.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.