Precision COPD diagnostics via non-invasive VOC-based breathomics and multi-omics integration

Background: Respiratory diseases are prevalent, frequently debilitating, and impose a substantial public health, medical, and economic burden. Chronic obstructive pulmonary disease (COPD) is a leading cause of morbidity and mortality globally, with its prevalence in Taiwan increasing from 2.5% to 6.1% in two decades (1996 to 2015). The severity, frequency of exacerbations, and response to treatment in COPD demonstrate considerable variability. Although spirometry remains the gold standard for diagnosis, its limited sensitivity and practicality in outpatient settings, particularly when patients present with the nonspecific symptom of “dyspnea,” underscore the urgent need for noninvasive, rapid diagnostic tools that can be conveniently implemented in clinical care. In our study, we utilize breathomics, specifically volatile organic compound (VOC) profiling, which provides a promising solution. Methods: We have enrolled 226 patients. The cohort investigates metal exposure, allergic profiles, and tailored pulmonary rehabilitation in COPD patients. This transdisciplinary design enables multi-omics-supported AI modeling for early diagnosis and personalized therapy. In case-control and case-crossover study designs, we collect and analyze VOC signatures in relation to clinical indicators such as FEV 1 , mMRC, CAT, exacerbation history, and responses to medications including ICS/LABA/LAMA. A hybrid material combining boron carbon nitride (BCN) and metal–organic frameworks (MOF) collect VOCs from breath, which are later analyzed using GC–MS. This provides a noninvasive way to detect breath biomarkers for health diagnostics. The goal is to integrate breathomics, multi-omics data, and clinical parameters using AI/ML modeling. Machine learning approaches (LASSO, Random Forest, XGBoost, Deep Learning) are employed to construct predictive models and incorporate VOC and multi-omics markers identified from the clinical data of the study cohort. Results: We collected PM2.5 heavy metal components from the air and measured cytokine levels in patients’ blood and urine. Urinary copper (Cu), selenium (Se), and molybdenum (Mo) collectively influence COPD development. Subsequently, we developed an artificial intelligence-driven clinical decision support system (CDSS) that integrates breathomics with multi-omics and clinical data. Mechanistic anchors, such as 4-HNE- and lipid-related metabolites, help ensure that the predictive models are biologically interpretable rather than solely statistically relevant. Conclusion: The validation of breathomics-AI as a swift and non-invasive instrument for the differential diagnosis of dyspnea is essential. Interventional evidence will substantiate the responsiveness of VOC signatures to therapeutic interventions, thereby endorsing their utility for both diagnostic and treatment-monitoring purposes. We have developed a clinically translatable pipeline from breath sampling to AI-driven decision support, which enhances precision in COPD care and offers a framework applicable to other respiratory conditions. This abstract was presented at the American Physiology Summit 2026 and is only available in HTML format. There is no downloadable file or PDF version. The Physiology editorial board was not involved in the peer review process.