Exhaled VOCs can discriminate subjects with COVID-19 from healthy controls.

ABSTRACT

COVID-19 detection currently relies on testing by reverse transcription polymerase chain reaction (RT-PCR) or antigen testing. However, SARS-CoV-2 is expected to cause significant metabolic changes in infected subjects due to both metabolic requirements for rapid viral replication and host immune responses. Analysis of volatile organic compounds (VOCs) from human breath can detect these metabolic changes and is therefore an alternative to RT-PCR or antigen assays. To identify VOC biomarkers of COVID-19, exhaled breath samples were collected from two sample groups into Tedlar bags negative COVID-19 (n= 12) and positive COVID-19 symptomatic (n= 14). Next, VOCs were analyzed by headspace solid phase microextraction coupled to gas chromatography-mass spectrometry. Subjects with COVID-19 displayed a larger number of VOCs as well as overall higher total concentration of VOCs (p< 0.05). Univariate analyses of qualified endogenous VOCs showed approximately 18% of the VOCs were significantly differentially expressed between the two classes (p< 0.05), with most VOCs upregulated. Machine learning multivariate classification algorithms distinguished COVID-19 subjects with over 95% accuracy. The COVID-19 positive subjects could be differentiated into two distinct subgroups by machine learning classification, but these did not correspond with significant differences in number of symptoms. Next, samples were collected from subjects who had previously donated breath bags while experiencing COVID-19, and subsequently recovered (COVID Recovered subjects (n= 11)). Univariate and multivariate results showed >90% accuracy at identifying these new samples as Control (COVID-19 negative), thereby validating the classification model and demonstrating VOCs dysregulated by COVID are restored to baseline levels upon recovery.