Breathomics-Guided Solid-State Sensor for Noninvasive Point-of-Care Diabetes Screening

Diabetes mellitus (DM) screening remains a major global health concern, with over 60% of cases undiagnosed. While breath analysis offers a noninvasive alternative to blood-based methods, current studies focus solely on acetone and lack diagnostic specificity. Here, we report an integrated diagnostic strategy combining metabolomics-driven biomarker discovery, portable solid electrolyte gas sensors (SEGS), and cellular-level metabolic investigation. Using GC-MS profiling of 130 DM patients and 122 healthy controls, we identified nine discriminative volatile organic compounds (VOCs). A random forest (RF) model achieves a cross-validated AUC of 0.93. The SEGS analyzer detects target VOCs at ppb levels within 30 s, enabling point-of-care (POC) screening. Clinical validation demonstrates 100% accuracy (n = 10) for diabetic ketoacidosis (DKA) detection and 83.3% (n = 30) for DM. Insulin-resistant (IR) cell models uncover breath VOCs' association with nonvolatile metabolite (NVM) pathways, supporting biological interpretability. This work establishes a biologically interpretable, clinically validated, and field-deployable diagnostic platform for scalable, low-cost DM screening in community and resource-limited settings.