Peripheral biochemical changes in major depressive disorder using breathomics analysis : new insights into molecular psychiatry

Major depressive disorder (MDD) is one of the most prevalent and disabling psychiatric illnesses worldwide, contributing significantly to the global burden of disease. Despite extensive research, no single biological mechanism fully explains its heterogeneity. Moreover, the clinical management of MDD remains challenging due to limited therapeutic success, imprecise subjective diagnostic criteria, and an incomplete understanding of the underlying molecular pathology. The metabolome represents the integrated output of complex interactions among the genome, transcriptome, proteome, microbiome, and environmental exposome. From this system-level perspective, we hypothesized that breath-based metabolomics, or breathomics, could non-invasively capture distinct, multi-parametric metabolic signatures associated with the heterogeneous pathophysiology of MDD. The primary aim of this study was to investigate the potential and limitations of the breath gas analysis in identifying known alterations in metabolic pathways implicated in MDD and to support the development of an objective, non-invasive, systems-level molecular diagnostic test, thereby contributing to the advancement of precision psychiatry. Particular attention was paid to post-awakening metabolite dynamics, given the critical role of this time window in activating stress-responsive biological systems such as hypothalamic–pituitary–adrenal axis. A total of 153 breath samples were collected at awakening, 30- and 60-minutes post awakening from 26 patients with MDD and 25 HCs. Exhaled volatile organic compounds (VOCs) were profiled using the proton transfer reaction mass spectrometry (PTR-MS). Candidate metabolites were initially preselected using the Wilcoxon tests and Student’s t-tests. Metabolomic pathway analysis and hierarchical clustering analysis (HCA) were then performed to group significant VOCs according to shared temporal dynamics and biological functions. Neighborhood component analysis was applied to rank VOCs according to their importance in classifying MDD and HCs. Forward feature selection was used as an iterative procedure to build multivariable predictive models and identify potential breathprint combinations. Four machine learning models: support vector machine (SVM), kernel SVM (K-SVM), Naïve Bayes (NB), and logistic regression (LR), were employed to evaluate the classification performance of individual VOCs and breathprints. Model accuracy was validated through Bootstrapped receiver operating characteristics (ROC) analysis. In total, 26 exhaled VOCs were significantly altered in MDD patients compared to HCs. These VOCs were enriched in five key metabolic pathways, including valine, leucine and isoleucine leucine biosynthesis (p = 5.50 10-4), pyruvate metabolism (p = 7.97 10-4), glycolysis and gluconeogenesis (p = 0.0015), nicotinate and nicotinamide metabolism (p = 0.005) and valine, leucine and isoleucine degradation (p = 0.005). HCA revealed distinct VOC clusters based on their biological functions, reflecting the interaction between host metabolome, exposome, and microbiome. Notably, VOCs of microbial origin exhibited the most pronounced alterations through various fermentation and metabolic pathways. Moreover, the combination of methylpyridine, isoprene, butylamine at their most significant awakening response times represents the most promising 3-variable breathprint across all models (AUC > 0.85, Sensitivity > 0.80, Specificity > 0.85). The addition of trimethylamine further improved model performance, while expanding to five-marker signatures by incorporating propionic acid or benzene further enhanced specificity, supporting their utility for phenotype-specific diagnostics. Furthermore, substituting components of the core breathprint VOCs with similar biological functions yielded comparable high diagnostic performance, suggesting metabolic redundancy, and offering flexibility in breathprint design. This study provides proof-of-concept that breathomics offers novel insights into the peripheral biochemical alterations associated with MDD, and holds strong potential for developing robust, non-invasive molecular breath signatures capable of distinguishing MDD from HCs with clinically meaningful accuracy. By capturing the multifaceted metabolic alterations underlying MDD and reflecting the complex interactions between host metabolome, gut microbiome and exposome, breathomics supports systems-level diagnostics and aligns with the goals of precision medicine in psychiatry. Notably, the identified breath biomarkers and associated pathways demonstrated strong concordance with findings from previous metabolomics studies using other biomatrices, further reinforcing their biological validity and translational relevance.