New Sequencing Test Shows Strong Accuracy in Spotting Drug-Resistant Tuberculosis

Tuberculosis remains one of the deadliest infectious diseases in the world, and its growing resistance to antibiotics is one of the most stubborn obstacles facing global health. The standard tools used to detect resistance in Mycobacterium tuberculosis, the bacterium that causes TB, are often slow, technically demanding, or limited in the number of drugs they can assess. Now, a team of researchers in Uganda has put a next-generation sequencing assay through one of the most rigorous independent evaluations it could receive, testing it against a panel of well-characterized bacterial isolates drawn from an international proficiency testing program. Their findings, published in BMC Genomics, offer both reassurance and a clear-eyed warning about the limits of even the most advanced molecular diagnostics.

The assay under scrutiny is Deeplex Myc-TB, a targeted next-generation sequencing test developed to detect mutations in the tuberculosis genome that are associated with resistance to a broad range of first- and second-line anti-TB drugs. Unlike conventional rapid molecular tests, which typically interrogate only a handful of genetic regions, targeted sequencing approaches can survey dozens of genes and intergenic regions in a single run. This breadth matters enormously in clinical practice, because treatment decisions for multidrug-resistant and extensively drug-resistant tuberculosis depend on knowing, quickly and reliably, which drugs are likely to fail and which are likely to work.

To evaluate the assay, the researchers turned to an unusual but powerful resource: isolates distributed through the World Health Organization global Tuberculosis Drug-Susceptibility Testing Proficiency Testing Program, organized and administered by the TB Supranational Reference Laboratory in Antwerp, Belgium. A total of 41 isolates were included in the study. These samples are particularly valuable for verification work because they have been exhaustively characterized by reference laboratories using multiple independent methods. In this study, phenotypic drug susceptibility testing performed on Löwenstein-Jensen medium and in mycobacteria growth indicator tube culture, together with whole-genome sequencing, served as the reference standards against which the Deeplex assay was judged.

The analytical workflow was straightforward in concept. Remnant DNA extracted from the proficiency testing isolates was subjected to targeted next-generation sequencing using the Deeplex Myc-TB assay. The resulting sequence data were interpreted using the manufacturer’s Deeplex web application, and, in parallel, with TBProfiler, an open-source tool for predicting drug resistance from mycobacterial sequence data, applying a composite interpretation rule. Concordance, sensitivity, and specificity were then calculated for each drug class, providing a detailed picture of where the assay excels and where it falls short.

The headline result is striking: the Deeplex Myc-TB assay achieved 100 percent specificity across all drug targets examined. In diagnostic terms, this means that whenever the assay reported a resistance mutation, the isolate genuinely carried a mutation associated with resistance. False positives, which can lead clinicians to withhold effective drugs unnecessarily, were entirely absent from the dataset. For several drugs, the assay’s sensitivity was equally impressive. It correctly identified resistance in every tested isolate for rifampicin, detecting 17 of 17 resistant strains; for pyrazinamide, 7 of 7; for fluoroquinolones, 13 of 13; and for linezolid, 6 of 6. For isoniazid, sensitivity was 94.1 percent, with 16 of 17 resistant isolates correctly flagged.

The single isoniazid miss is instructive rather than alarming. The one false-susceptible result traced back to a specific insertion mutation in the katG gene, designated katG c.45_46insA, which fell outside the region targeted by the assay. This illustrates a fundamental property of targeted sequencing: its performance is bounded by the mutations it is designed to detect. When a resistance-conferring variant lies outside the panel’s target regions, even a technically flawless assay will report susceptibility. The finding underscores why laboratories adopting such assays must understand the local epidemiology of resistance mutations in the strains they encounter.

Performance was more mixed for the newer and repurposed drugs that have become central to modern regimens for drug-resistant tuberculosis. For clofazimine, the assay detected resistance in 4 of 5 phenotypically resistant isolates, a sensitivity of 80 percent. For bedaquiline, sensitivity dropped to 44.4 percent, with only 4 of 9 resistant strains correctly identified. In five cases, isolates that were resistant by phenotypic testing carried no resistance variants detected by the assay. The authors traced much of this gap to the biology of resistance to these drugs. Mutations in the mmpR5 gene, also known as Rv0678, a major regulator of efflux pump expression, were largely detected by the assay, but resistance mediated through the atpE gene was not identified, because that gene is not adequately covered by the assay’s target panel.

Overall concordance between the Deeplex assay and the reference standards ranged from 87.5 percent for bedaquiline to a perfect 100 percent for rifampicin, pyrazinamide, fluoroquinolones, and linezolid. Notably, the assay consistently detected several recurrent mutations associated with resistance to the different drugs, reinforcing the idea that for the common, well-established resistance mechanisms, targeted sequencing performs at or near the level of whole-genome sequencing while being faster and cheaper to deploy in routine laboratory settings. For a disease where every week of delayed or inappropriate treatment increases the risk of transmission, disease progression, and death, that speed advantage carries real clinical weight.

The implications of this verification study extend well beyond one laboratory in Uganda. Proficiency testing panels offer a rare opportunity to benchmark a diagnostic assay against isolates whose resistance profiles are known with high confidence, free from the uncertainties that complicate evaluations using routine clinical samples. By demonstrating near-perfect specificity and strong sensitivity for the cornerstone drugs of tuberculosis therapy, the study provides laboratories considering adoption of targeted next-generation sequencing with concrete, independently grounded performance data. At the same time, the documented gaps for bedaquiline and clofazimine serve as a caution against treating any single assay as a complete solution. The authors emphasize that understanding local resistance mutation patterns is essential when implementing sequencing-based resistance detection, because an assay that misses the mutations circulating in a given region will systematically underreport resistance there.

The broader context makes these findings timely. Drug-resistant tuberculosis kills thousands of people each year, and the introduction of bedaquiline, linezolid, and other newer agents has transformed outcomes for patients with resistant disease, but only when those drugs are deployed appropriately. Molecular diagnostics that can rapidly and accurately profile resistance across the full range of current and repurposed drugs are a critical link in that chain. This study shows that targeted sequencing platforms have matured to the point where they can match reference standards for the most important drugs, while also mapping precisely where the remaining blind spots lie. For laboratory directors, national tuberculosis programs, and diagnostic developers alike, the message is twofold: the technology is ready for prime time in the core of TB drug resistance testing, but careful attention to target coverage, and to the mutation landscape of the populations being served, remains indispensable.

Subject of Research: Analytical performance verification of a targeted next-generation sequencing assay for detecting drug-resistant Mycobacterium tuberculosis

Article Title: Analytical performance verification of the Deeplex Myc-TB assay using well-characterized Mycobacterium tuberculosis proficiency testing isolates

Article References: Kabahita, J. M., Lunkuse, J. M., Batte, D. N., Nakato, H. S., Namutebi, J., Kabugo, J., Adam, I., Kanyerezi, S., Makoha, C., Nsubuga, M. L., Oundo, H. R., Wenka, G. T., Sseruyange, J., Murungi, M., Kasule, G. W., Lutaaya, P., Kyokushaba, J., Byabajungu, H., Ademun, P., … Joloba, M. L. (2026). Analytical performance verification of the Deeplex Myc-TB assay using well-characterized Mycobacterium tuberculosis proficiency testing isolates. BMC Genomics. https://doi.org/10.1186/s12864-026-13336-z

Image Credits: AI Generated

DOI: 10.1186/s12864-026-13336-z

Keywords: tuberculosis, drug resistance, Deeplex Myc-TB, targeted next-generation sequencing, whole-genome sequencing, diagnostics, bedaquiline, clofazimine, isoniazid, rifampicin, proficiency testing, Uganda