Rapid, non-invasive diagnosis of tuberculosis using desorption separation ionization mass spectrometry (DSI-MS) and urinary carbonyl metabolite fingerprints.

Tuberculosis (TB) remains a leading global infectious disease that demands rapid, non-invasive diagnostic solutions. Here, we present a rapid urine-based diagnostic platform that integrates desorption separation ionization mass spectrometry (DSI-MS) with Amine reagents N-(2-Aminoethyl) piperidine (AED) carbonyl derivatization to enhance detection of urinary carbonyl metabolites. Minimal sample pretreatment was required: 100 μL urine samples were mixed with 2 μL AED and an 10 μL internal standard (9-phenylacridine), heated (70 °C, 5 min), applied to a sand-core filter and analyzed on an LTQ Orbitrap Velos Pro coupled to the DSI device (heater 230 °C; positive-ion acquisition m/z 50-500; acquisition time 3 min). Using this approach, we identified 65 metabolites (32 derivatized and 33 underivatized) in urine samples from a cohort of 151 tuberculosis patients and 151 healthy controls, with samples collected at the First Affiliated Hospital of Gannan Medical University and Beijing Hospital. Derivatization substantially increased ionization efficiency for carbonyl-containing metabolites and expanded coverage of low-abundance species. Pathway analysis highlighted perturbation of energy- and nitrogen-related metabolism, notably arginine and proline metabolism with elevated ornithine in TB patients. A neural-network classifier trained on the combined metabolite fingerprint achieved high discriminative performance (AUC = 0.927 on the training set and AUC = 0.922 on the test set). These results demonstrate that DSI-MS with AED derivatization, combined with machine learning, provides a rapid, non-invasive approach for TB screening and yields metabolic insights warranting further validation.