CO2-Gated End-Tidal Breathomics Maps Gastrointestinal Malignant Risk

Abstract

Human breath volatile organic compounds (VOCs) reflect alterations in cellular metabolism, inflammatory responses and oxidative stress, providing potential molecular readouts for noninvasive disease monitoring. However, conventional breath sampling commonly relies on whole-breath collection, which readily mixes with oral and upper-airway dead-space air, diluting disease-associated VOC signals that originate from alveolar exchange and limiting its application in noninvasive diagnosis. To address this limitation, we developed a CO2-gated intelligent end-tidal breath sampler (iEBS) based on dynamic changes in end-tidal CO2 partial pressure and coupled it with TD-GC-MS/MS to establish an end-tidal breathomics platform for VOC profiling. Using breath profiles from 273 participants, including healthy controls and individuals with gastritis, polyps, gastric cancer and colorectal cancer, we reproducibly identified and quantified 73 VOCs. By integrating topological data analysis, multistage risk modeling, chemical structure enrichment and single-cell transcriptomic mapping, we constructed a visualization framework for gastrointestinal malignancy risk. This framework revealed a gradient of volatile metabolic remodeling from early gastrointestinal abnormalities to malignant states and nominated propanal, butyric acid and 1-butanol as representative VOC markers with potential mechanistic relevance. The multistage risk model achieved stable disease stratification performance (AUC = 0.78–0.94), and a proof-of-concept pre-endoscopic triage simulation estimated that the average waiting time for malignant cases could be reduced by 20%. Overall, this study establishes an iEBS-based breath VOC profiling approach for phase-controlled end-tidal collection, reproducible quantification and interpretable visualization of alveolar-derived breath signals, providing an analytical framework for the noninvasive assessment of gastrointestinal diseases.