Although the main components of human breath are nitrogen, carbon dioxide, oxygen, and water vapor it also contains over 100 other gases such as hydrogen, methane, and VOCs [13]. After the exhaled gas is collected in a bag, it is adsorbed and stored in a special collection tube, which has been improved to enable detection of subject-specific volatile substances [14, 15]. While the trace substances detected in exhaled gas have individual-specific patterns, there were no gender or age differences. Since substances specific to each individual can be stably quantified, this method is useful as a clinical test.
We could distinguish between substances that undergo specific changes due to inflammation and those that do not. In diseases such as asthma, lipid mediators (Leukotriene C4/D4/E4), oxidative stress markers (H2O2, 8-isoprostane), cytokines and chemokines (IL-4, 8, 17, TNF-α, RANTES, IP-10, TGF-β, MIP-1α β) are increased in exhaled gas condensate [16, 17]. Hence we hypothesized that trace component analysis of exhaled gas would further elucidate the pathogenesis of inflammatory diseases. To date, studies on exhaled VOCs have identified biomarkers for diseases such as lung cancer [18–21], breast cancer [22], and pulmonary tuberculosis [23, 24]. In this study, exhaled gas was collected in a group of patients with non-infectious inflammatory diseases, including rheumatoid arthritis and polymyalgia rheumatica, during the inflammatory phase before treatment. Cluster analysis of these samples showed that class prediction was generally possible. In addition, PCA analysis, in which the ESR and a large amount of exhaled gas data were compressed to three dimensions, showed that the data for non-inflammatory conditions were arranged on a specific surface, and discrimination was therefore possible. In some patients, exhaled gas was collected when the inflammation had stabilized after treatment. The exhaled gas pattern showed either an increase or decrease in the presence of inflammation and an unidentified substance that varied with treatment. In this study, it was not confirmed whether the normalization of this substance to healthy levels were solely due to treatment. However, the observation of such a substance is a significant finding because of its potential as a clinical biomarker. In this study, the substances were not identified because they were not analyzed by a test system connected to a mass spectrometer and was only estimated from an industrial library. Moreover, some substances were observed to change as the inflammatory state fluctuated, but it is possible that they were derived from the same larger molecule in the original formation process and degraded in the metabolic process. Even substances produced by metabolism, including the original substances, cannot be detected as biogas if they are non-volatile. In future, we will consider further analysis using a mass spectrometer to identify the individual substances. There are many substances that can sensitively reflect minute metabolic changes, and some of them are new substances. The novel substances have not been identified using the latest version of the data library. For this reason, we are currently trying to increase the exhalation volume and the number of samples. Although modern exhaled gas assays with highly selective technology can detect over 2,000 different biomarkers in a single sample, the large number of candidates results in an increased statistical risk of identifying a substance as a disease biomarker when it is not [14, 25]. To minimize detection of false positive markers, it is necessary to derive and validate a training set that uses only those biomarkers identified with high statistical significance. It must be viewed as a group of substances showing a pattern change in the content of exhaled gas components. Therefore, if we want to analyze changes in a large number of substances, we need to make decisions based on an accumulation of the results of multiple analyses, and decision analysis of a large number of breath components by artificial intelligence (AI) is necessary. It may be possible to develop new metabolic research such as prognosis prediction by combining ultra-early diagnosis of the disease state with AI. Multiomics-related machine learning shows promise for the assessment of heart failure [26]. Diagnostic sensitivity and specificity can be improved by increasing case numbers in the future.
Since exhaled gas measurement is non-invasive, it can be conducted repeatedly and even in non-medical facilities. Therefore, this test is considered significant since patients could potentially use it to monitor and maintain their condition. The method presented in this study can be a comprehensive method for identifying disease-specific biogas markers including inflammatory diseases and other systemic diseases. However, with the spread of infectious diseases such COVID-19, the possible contamination of specimen samples with viruses is a pressing concern, and it is necessary to confirm that there is no infection during sample collection. Research has been conducted to determine the presence or absence of SARS-CoV-2 infection by exhaled gas, and the non-invasiveness and simplicity of exhaled gas analysis is expected to be applied to exhaled gas omics analysis by expanding the analysis of exhaled gas to metabolites, viruses, and genes [27, 28].