BackgroundMajor depressive disorder (MDD) is one of the most common psychiatric disorders with multifactorial etiologies. Metabolomics has recently emerged as a particularly potential quantitative tool that provides a multi-parametric signature specific to several mechanisms underlying the heterogeneous pathophysiology of MDD. The main purpose of the present study was to investigate possibilities and limitations of breath-based metabolomics, breathomics patterns to discriminate MDD patients from healthy controls (HCs) and identify the altered metabolic pathways in MDD.MethodsBreath samples were collected in Tedlar bags at awakening, 30 and 60 min after awakening from 26 patients with MDD and 25 HCs. The non-targeted breathomics analysis was carried out by proton transfer reaction mass spectrometry. The univariate analysis was first performed by T-test to rank potential biomarkers. The metabolomic pathway analysis and hierarchical clustering analysis (HCA) were performed to group the significant metabolites involved in the same metabolic pathways or networks. Moreover, a support vector machine (SVM) predictive model was built to identify the potential metabolites in the altered pathways and clusters. The accuracy of the SVM model was evaluated by receiver operating characteristics (ROC) analysis.ResultsA total of 23 differential exhaled breath metabolites were significantly altered in patients with MDD compared with HCs and mapped in five significant metabolic pathways including aminoacyl-tRNA biosynthesis (p = 0.0055), branched chain amino acids valine, leucine and isoleucine biosynthesis (p = 0.0060), glycolysis and gluconeogenesis (p = 0.0067), nicotinate and nicotinamide metabolism (p = 0.0213) and pyruvate metabolism (p = 0.0440). Moreover, the SVM predictive model showed that butylamine (p = 0.0005, pFDR=0.0006), 3-methylpyridine (p = 0.0002, pFDR = 0.0012), endogenous aliphatic ethanol isotope (p = 0.0073, pFDR = 0.0174), valeric acid (p = 0.005, pFDR = 0.0162) and isoprene (p = 0.038, pFDR = 0.045) were potential metabolites within identified clusters with HCA and altered pathways, and discriminated between patients with MDD and non-depressed ones with high sensitivity (0.88), specificity (0.96) and area under curve of ROC (0.96).ConclusionAccording to the results of this study, the non-targeted breathomics analysis with high-throughput sensitive analytical technologies coupled to advanced computational tools approaches offer completely new insights into peripheral biochemical changes in MDD.