Abstract
Tuberculosis (TB) has high morbidity and mortality rates, and drug-resistant strains pose an increasing challenge. Traditional methods for detecting the Mycobacterium tuberculosis complex (MTBC) are insufficient for rapid clinical diagnosis. This prospective study compared the diagnostic efficacy of targeted next-generation sequencing (tNGS), metagenomic next-generation sequencing (mNGS), and Xpert MTB/RIF using bronchoalveolar lavage fluid (BALF) samples from 121 patients with suspected pulmonary TB. Against the reference standard of mycobacterial culture, tNGS demonstrated the highest sensitivity (97.44%), followed by Xpert MTB/RIF (92.31%) and mNGS (84.62%), specificities were 69.51%, 69.51%, and 75.61%, respectively. To address the limitations of culture as an imperfect reference standard and the potential bias from clinical diagnosis, Bayesian Latent Class Analysis (BLCA) was employed. BLCA, which does not assume a perfect gold standard, estimated sensitivities of 98.7%, 99.8%, and 91.0% for tNGS, Xpert MTB/RIF, and mNGS, with corresponding specificities of 89.3%, 93.3%, and 97.8%, respectively. Both tNGS and Xpert MTB/RIF consistently detected rifampicin resistance mutations (rpoB) (p = 0.219, Kappa = 0.730). In conclusion, tNGS offers comparable specificity and sensitivity to Xpert MTB/RIF for TB diagnosis, with the advantage of distinguishing between MTBC, non-tuberculous Mycobacteria (NTM), and other microorganisms. Simultaneously, it provides insights into anti-TB drug resistance. Thus, tNGS is a valuable tool for diagnosing TB in various clinical settings. Clinical trial number, Not applicable. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12879-026-12673-4.