Modelling electronic nose sensor deflections by matching Gas Chromatography-Mass Spectrometry exhaled breath samples

Abstract

Rationale: Analysis of exhaled volatile organic compounds (VOCs) by sensor-driven electronic nose (eNose) technology is a widely suggested measure for non-invasive monitoring of chronic airway diseases. While this technology allows probabilistic (clinical) advise, it is incapable of providing details regarding involved metabolites. Modelling sensor deflections by matching Gas Chromatography-Mass Spectrometry (GC-MS) samples could help to ascertain which VOCs induce a sensor response. Objective: To determine the association between eNose sensor deflections and exhaled VOCs measured by GC-MS. Methods: Paired samples of breath from asthma patients (n=22) were collected on sorbent tubes. One tube was analysed by four different eNoses and the second one by GC-MS. Pooling of resulting datasets consisted of 1) Partial Least Square Regression (PLSR) analysis and 2) clustered heat-mapping of PLSR loadings. Results: Matching data was available for 158 eNose sensors and 1025 GC-MS features, whereby PLSR resulted in R2‘s from 0.25 to 0.72. Figure 1 shows clustered heat-mapping outcomes. Conclusion: This explorative analysis revealed distinctive and associated patterns of exhaled VOCs between eNose and GC-MS. This data matching could help to identify which VOCs are responsible for clinically relevant results obtained by eNose and will facilitate valorisation of breathomics into clinical tests.