Unbiased clustering of severe asthma patients based on exhaled breath profiles

Abstract

Rationale: Severe asthma is a heterogeneous clinical condition including various pathophysiological pathways. Metabolomics of exhaled air is associated with airways inflammation in patients with asthma and COPD. Aim: To reveal severe asthma phenotypes by unbiased cluster analysis based on metabolomic fingerprints from exhaled breath by gas-chromatography/mass-spectrometry (GCMS) and to link results to electronic nose (eNose) data. Methods: This was a cross-sectional analysis in the U-BIOPRED cohort. Severe asthma was defined by IMI-criteria [Bel Thorax 2011]. Exhaled volatile organic compounds (VOCs) were trapped on two adsorption tubes per subject: one for GCMS and one for the U-BIOPRED eNose platform (Owlstone Lonestar, Cyranose C320, Comon Invent, Tor Vergata TEN). Data cleaning on both omic datasets included: normalization and data reduction by principal component analysis (PCA). Statistical analysis of GCMS data was performed using Ward clustering, followed by Similarity Profile Analysis. The between-cluster comparison of baseline variables and eNose PCs was done by ANOVA, Kruskal-Wallis or χ2 tests. Results: Shared GCMS-eNose data were available for 35 patients (age 50±15yr, 45% male, 46% (ex-)smokers). Four clusters of GCMS data were delineated, that differed significantly regarding: SNOT questionnaire (p=0.02), smoking pack years (p=0.04), sputum neutrophils (p=0.04), serum periostin (p=0.03) and 7 out of 12 eNose PCs (p Conclusions: Unbiased fingerprinting of exhaled air by GCMS provides clusters of severe asthma patients that differ regarding clinical and inflammatory parameters. Breath analysis by eNose technology is resembling, but not identical to GCMS in severe asthma.