Plasma miRNA signature for the diagnosis of pulmonary tuberculosis in symptomatic patients

BACKGROUND: The prompt and precise diagnosis of active pulmonary tuberculosis (TB) is crucial for controlling this disease and yet it remains a global challenge. The objective of this study was to identify a set of microRNAs (miRNAs) whose expression in plasma could be used as a triage test for diagnosing TB. METHODS: A total of 879 plasma samples were collected in seven clinical centres from healthy individuals and patients displaying TB-like symptoms and/or radiological features consistent with TB. The samples were classified as TB, pneumonia, lung cancer and HC subgroups based on subsequent diagnostic assessments.We performed quantitative profiling of 264 plasma miRNAs in a training cohort (n=410) and an independent external test cohort (n=469). After dimensionality reduction and feature selection analysis, we identified nine discriminative miRNAs and used them to train an ensemble model in a training cohort using the scikit-learn library, which was subsequently evaluated in the external test cohort. RESULTS: The ensemble model showed notable accuracy in discriminating TB from non-TB patients, yielding areas under the curve (AUC) of 0.84 (95% CI 0.80 to 0.88) for the training cohort and 0.86 (95% CI 0.82 to 0.90) for the external test cohort. When tested against subgroups of laboratory confirmed and clinically diagnosed but unconfirmed TB, the AUC values were 0.89 (95% CI 0.85 to 0.93) and 0.83 (95% CI 0.79 to 0.88), respectively. In smear-negative confirmed TB patients, the AUC exceeded 0.83, with a sensitivity and specificity of 0.75. CONCLUSIONS: Our miRNA ensemble model, based on detecting a nine-miRNA expression biosignature in plasma, demonstrated promising ability to diagnose TB and distinguish it from other common lung diseases but further studies are needed to assess its clinical applicability. TRIAL REGISTRATION NUMBER: ChiCTR2000039734.