Premium
Understanding uncontrolled severe allergic asthma by integration of omic and clinical data
Author(s) -
DelgadoDolset María Isabel,
Obeso David,
RodríguezCoira Juan,
Tarin Carlos,
Tan Ge,
Cumplido José A.,
Cabrera Ana,
Angulo Santiago,
Barbas Coral,
Sokolowska Milena,
Barber Domingo,
Carrillo Teresa,
Villaseñor Alma,
Escribese María M.
Publication year - 2022
Publication title -
allergy
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.363
H-Index - 173
eISSN - 1398-9995
pISSN - 0105-4538
DOI - 10.1111/all.15192
Subject(s) - metabolomics , asthma , medicine , disease , proteomics , allergy , immunology , bioinformatics , biology , biochemistry , gene
Background Asthma is a complex, multifactorial disease often linked with sensitization to house dust mites (HDM). There is a subset of patients that does not respond to available treatments, who present a higher number of exacerbations and a worse quality of life. To understand the mechanisms of poor asthma control and disease severity, we aim to elucidate the metabolic and immunologic routes underlying this specific phenotype and the associated clinical features. Methods Eighty‐seven patients with a clinical history of asthma were recruited and stratified in 4 groups according to their response to treatment: corticosteroid‐controlled (ICS), immunotherapy‐controlled (IT), biologicals‐controlled (BIO) or uncontrolled (UC). Serum samples were analysed by metabolomics and proteomics; and classifiers were built using machine‐learning algorithms. Results Metabolomic analysis showed that ICS and UC groups cluster separately from one another and display the highest number of significantly different metabolites among all comparisons. Metabolite identification and pathway enrichment analysis highlighted increased levels of lysophospholipids related to inflammatory pathways in the UC patients. Likewise, 8 proteins were either upregulated (CCL13, ARG1, IL15 and TNFRSF12A) or downregulated (sCD4, CCL19 and IFNγ) in UC patients compared to ICS, suggesting a significant activation of T cells in these patients. Finally, the machine‐learning model built including metabolomic and clinical data was able to classify the patients with an 87.5% accuracy. Conclusions UC patients display a unique fingerprint characterized by inflammatory‐related metabolites and proteins, suggesting a pro‐inflammatory environment. Moreover, the integration of clinical and experimental data led to a deeper understanding of the mechanisms underlying UC phenotype.