Rapid Sequencing Outperforms Routine Tests in Detecting Tuberculosis, Study Finds

Metagenomic next-generation sequencing, a technique that can identify genetic fragments of any pathogen in a clinical sample without needing to know in advance what to look for, detected Mycobacterium tuberculosis far more often than the standard battery of tuberculosis tests in a large retrospective study conducted at Sichuan Provincial People’s Hospital in Chengdu, China. The findings, published in Frontiers in Cellular and Infection Microbiology, suggest that the technology could serve as a valuable adjunctive tool for clinicians facing difficult mycobacterial diagnoses, particularly when conventional tests come back unrevealing or indeterminate.

The research team, led by Xiaoli Zhou, Kai Zhang, and Aibo Liu of the Department of Laboratory Medicine, reviewed the records of 425 patients with suspected tuberculosis infection who underwent sequencing-based testing at the hospital between 2021 and 2024. Of these, 306 patients, or 72 percent, were ultimately classified as having pulmonary tuberculosis, 67 patients, or 15.76 percent, had extrapulmonary tuberculosis, and 52 patients, or 12.24 percent, did not have tuberculosis at all. The final classifications were made according to the Chinese Clinical Treatment Guide for Tuberculosis and World Health Organization recommendations, drawing on microbiological evidence, radiological findings, clinical symptoms, response to anti-tuberculosis treatment, and exclusion of other possible diagnoses.

Because the study was retrospective, not all patients received every test. Results for the conventional methods, which included tuberculosis IgG antibody testing, the interferon-gamma release assay known as TB-IGRA, tuberculosis DNA detection by PCR, the GeneXpert MTB/RIF assay, mycobacterial culture, and acid-fast bacilli staining, were unavailable for between 221 and 356 patients depending on the method. Taking the final clinical diagnosis as the reference standard, sequencing achieved a sensitivity of 90.1 percent across all samples, with a 95 percent confidence interval of 0.866 to 0.927, and a specificity of 75.0 percent. That sensitivity exceeded those of TB-IGRA at 79.1 percent, Xpert at 72.4 percent, culture at 52.3 percent, TB-DNA at 45.6 percent, TB-IgG at 36.4 percent, and acid-fast staining at just 10.7 percent.

The specificity picture was more nuanced. Sequencing’s specificity of 75.0 percent was lower than that of Xpert and TB-DNA, both of which reached 100 percent in this cohort, and lower than TB-IgG at 91.2 percent and TB-IGRA at 88.6 percent, though it was higher than culture’s 65.9 percent. The positive predictive value of sequencing was 96.3 percent, but its negative predictive value was only 51.3 percent, meaning a negative result could not reliably rule out tuberculosis. The Youden index, a single measure combining sensitivity and specificity, was 65.1 percent for sequencing, which was actually lower than the values for Xpert at 72.4 percent and TB-IGRA at 67.7 percent, a reminder that higher sensitivity alone does not guarantee the best overall diagnostic trade-off.

Performance also varied by sample type. In 313 pulmonary cases, sequencing reached a sensitivity of 93.4 percent with a specificity of 81.0 percent. In 112 extrapulmonary cases, sensitivity fell to 81.4 percent and specificity dropped to 50.0 percent, whereas culture specificity in extrapulmonary samples climbed to 100 percent. The authors noted that microbiological methods performed particularly poorly on extrapulmonary specimens, likely because pathogen loads are lower outside the lungs, making culture and microscopy difficult. Imaging alone was also of limited help in many extrapulmonary cases; the study found that radiological findings were typical and easy to interpret for lumbar tuberculosis but challenging for tuberculosis involving other organs, especially when patients were co-infected with other pathogens.

A central finding concerned the technology’s ability to distinguish Mycobacterium tuberculosis from nontuberculous mycobacteria, a distinction that matters enormously because the two require entirely different treatments and are difficult to tell apart on imaging or routine laboratory results. Among the 52 non-tuberculosis patients, 39 were ultimately diagnosed with nontuberculous mycobacterial infection, and sequencing correctly identified nontuberculous mycobacteria in all 39 of them, yielding a sensitivity of 100 percent against the final clinical diagnosis and a positive predictive value of 97.5 percent. The authors cautioned, however, that with only one nontuberculous-negative case and no sequencing-negative results, specificity and negative predictive value for nontuberculous mycobacteria could not be reliably assessed, and the small sample size may have inflated the positive predictive value estimate.

The study also explored how the amount of pathogen in a sample shapes sequencing results. Because the waxy cell wall of M. tuberculosis is difficult to break open, DNA extraction can be inefficient, and sensitivity may suffer when bacterial burden is low. Among the 349 sequencing-positive samples, the number of MTB reads was significantly higher in patients who tested positive by each of the conventional methods, indicating that positive results on those tests reflect a high bacterial burden. The researchers also generated receiver operating characteristic curves for sequencing read counts and found an area under the curve of 0.9073 for MTB diagnosis, with a sensitivity of 73.19 percent and a specificity of 96.15 percent when a cutoff of 4 reads was applied, chosen by maximizing the Youden index at 0.6934. Notably, 35 sequencing-positive samples had only a single read mapped to MTB, and of these, 15 were confirmed positive by Xpert or culture while 10 were positive by sequencing alone, illustrating the interpretive difficulty posed by very low read counts.

Concordance analysis added further context. Among 270 patients tested simultaneously by sequencing, Xpert, and culture, all three methods agreed in only 116 patients, or 42.96 percent, with 45 samples positive by sequencing alone. Statistically, sequencing showed satisfactory concordance with Xpert across all samples and pulmonary samples, but poor concordance with culture, which the authors attributed partly to culture’s low sensitivity. Turnaround time was another advantage: because the sequencing platform was located within the hospital, results were available in 24 to 36 hours, compared with up to 42 days for mycobacterial culture, which remains the recognized gold standard because of its low cost and its ability to support drug susceptibility testing.

The authors were candid about the factors that may have inflated their sensitivity estimates. Using clinicians’ final diagnoses as the gold standard rather than microbiological confirmation alone, restricting enrollment to patients with clinically suspected tuberculosis whose imaging was consistent with the diagnosis, and a relatively small sample size may all have contributed to the elevated figure. Thirteen patients in the non-tuberculosis group had low-read-count sequencing results suggestive of potential MTB that were negative by culture, Xpert, and PCR, and these were counted as false positives. The authors emphasized that such results must be interpreted alongside clinical findings and conventional testing, and that confirmatory targeted assays or dynamic monitoring are recommended for organisms suspected to represent contamination or background microbiota.

The study had other limitations as well. It was single-center, not all patients underwent all tests because of the retrospective design, and the number of samples from sterile body fluids was small. The clinical benefits of sequencing, such as its impact on hospital stay and patient prognosis, remain to be investigated, and further research is needed to determine whether sequencing can distinguish active tuberculosis infection from latent infection. Still, the authors concluded that for patients with suspected tuberculosis or nontuberculous mycobacterial infection, sequencing may serve as a helpful adjunctive tool, particularly when routine diagnostic tests are unrevealing, and that its capacity to detect polymicrobial co-infections in a single assay adds another dimension to its clinical utility. They stressed that it should be viewed as a complement to, rather than a replacement for, conventional diagnostic methods.

Subject of Research: Diagnostic performance of metagenomic next-generation sequencing for tuberculosis and nontuberculous mycobacterial infection

Article Title: Diagnostic performance of metagenomic next-generation sequencing in Mycobacterium and non-Mycobacterium infection patients: a retrospective study at a large comprehensive medical center

Article References: Zhou, X., Peng, S., Tan, R., Wang, J., Hu, H., Zhang, K., & Liu, A. (2026). Diagnostic performance of metagenomic next-generation sequencing in Mycobacterium and non-Mycobacterium infection patients: a retrospective study at a large comprehensive medical center. Frontiers in Cellular and Infection Microbiology, 16, Article 1935536. https://doi.org/10.3389/fcimb.2026.1935536

Image Credits: AI Generated

DOI: 10.3389/fcimb.2026.1935536

Keywords: metagenomic next-generation sequencing, tuberculosis, Mycobacterium tuberculosis, nontuberculous mycobacteria, diagnostic sensitivity, Xpert MTB/RIF, mycobacterial culture, extrapulmonary tuberculosis, retrospective study, interferon-gamma release assay, acid-fast bacilli staining, China