TB-Profiler upgrade taps 171,000 genomes to sharpen drug-resistance prediction

Tuberculosis remains one of the deadliest infectious diseases on the planet, and the rise of drug-resistant strains has made rapid, accurate susceptibility testing a matter of life and death. Now a widely used open-source bioinformatics tool has received its most substantial upgrade yet. Writing in Genome Medicine, a team led by Jody E. Phelan and Taane G. Clark of the London School of Hygiene and Tropical Medicine describes an updated version of TB-Profiler that extends its drug-resistance predictions to 17 anti-tuberculosis drugs and plugs the platform into a curated global database of more than 171,000 Mycobacterium tuberculosis genomes drawn from 136 countries. The new release, version 6.6.5, is designed to serve both clinicians deciding how to treat an individual patient and public health officials tracking outbreaks across communities.

TB-Profiler was developed to profile M. tuberculosis directly from next-generation sequencing data. Rather than waiting days or weeks for laboratory cultures to grow, the tool scans raw sequencing reads for known resistance-conferring mutations and assigns the isolate to one of the major phylogenetic lineages of the tuberculosis bacillus. This approach, known as genotypic antimicrobial susceptibility prediction, has become increasingly important as sequencing costs fall and national tuberculosis programmes adopt whole-genome sequencing for routine diagnostics. The original version of the tool covered 16 anti-TB drugs; the updated release adds delamanid and pretomanid, two newer agents that are central to modern regimens for multidrug-resistant and extensively drug-resistant disease.

The technical heart of the upgrade lies in its expanded and refined mutation libraries. TB-Profiler does not attempt to predict resistance from first principles; instead, it relies on carefully curated catalogues of mutations that have been experimentally or epidemiologically linked to reduced drug susceptibility. The updated version incorporates interpretation rules endorsed by the World Health Organization, ensuring that the tool’s output aligns with internationally recognised standards for reporting resistance to drugs such as bedaquiline and clofazimine, cycloserine and terizidone, and para-aminosalicylic acid. The developers have also added curated loss-of-function mutations, which are changes that disable a gene entirely and are increasingly recognised as important drivers of resistance to newer drug classes.

Why does this matter clinically? Regimens for drug-resistant tuberculosis have changed dramatically in recent years, shifting towards shorter, all-oral combinations built around bedaquiline, pretomanid and linezolid. Choosing the right regimen requires knowing exactly which drugs the infecting strain can tolerate, and phenotypic antimicrobial susceptibility testing, the traditional gold standard, can take weeks because M. tuberculosis grows so slowly. Genotypic prediction from sequencing data can deliver a resistance profile in a fraction of that time, and tools like TB-Profiler make that analysis accessible to laboratories that lack deep bioinformatics expertise, through both a command-line version and a web interface.

The second major pillar of the update is population context. The team has integrated a continuously expanding global genomic database of more than 170,000 isolates, representing all major lineages of the M. tuberculosis complex and spanning 136 countries. This resource allows the tool to place any newly sequenced genome within its global population framework, comparing allele frequencies and identifying how unusual or common a given mutation is across the worldwide diversity of the pathogen. That context is technically important because not every mutation observed in a resistant isolate actually causes resistance; some are simply neutral variants that happen to be common in certain lineages. A global reference panel helps distinguish the two.

Built on top of this database is a new analytical capability for calculating and visualising genomic relatedness between isolates. In tuberculosis surveillance, genomic relatedness is the key to identifying transmission. When two isolates from different patients are nearly identical across their genomes, it strongly suggests that one infected the other, or that both were infected by a common source. By computing distances between genomes and presenting the results in accessible visualisations, the updated TB-Profiler allows public health teams to spot potential transmission chains and outbreaks directly within the same workflow they already use for resistance prediction, rather than juggling separate tools.

To demonstrate the new functionality, the researchers applied it to isolates from Uganda, where they were able to identify potential transmission events among the sampled genomes. The demonstration illustrates the practical value of combining resistance profiling and relatedness analysis in a single platform. In high-burden settings, distinguishing between reactivation of a latent infection picked up years ago and recent active transmission changes the public health response entirely: recent transmission calls for intensified contact tracing and community screening, whereas reactivation points towards the management of latent infection in vulnerable populations.

Accessibility has also been broadened in a way that could prove decisive for global uptake. The updated tool introduces multilingual reporting capabilities, meaning that the clinical reports generated from sequencing data can be delivered in the languages spoken by the health workers who use them. This may sound like a modest feature, but in tuberculosis control it addresses a real barrier: the countries carrying the greatest burden of drug-resistant disease are often those where English is not the primary working language of laboratory and clinical staff. By lowering the language barrier, the developers hope to widen implementation of genomic surveillance in precisely the regions where it is needed most.

The work is a product of an international collaboration spanning the London School of Hygiene and Tropical Medicine, Thailand’s Ministry of Public Health Medical Genetics Center, the South African National Bioinformatics Institute at the University of the Western Cape, Nottingham Trent University and the University of Cambridge. The informatics development was supported by a UKRI EPSRC award in artificial intelligence innovation to accelerate health research, with additional support from the UKRI Medical Research Council. The tool and its associated databases are available in both stand-alone and web-based versions at the team’s hosted platform, and the underlying article is published open access under a Creative Commons licence.

The developers are explicit that this release is a stepping stone rather than a destination. Future versions will leverage the ever-expanding sequencing database to deploy artificial intelligence approaches aimed at refining lineage assignment, improving drug-resistance prediction and sharpening transmission classification, including the identification of resistance-associated mutations that have not yet been characterised. As the global database grows with every sequenced isolate, the statistical power available to such machine-learning methods grows with it, promising a feedback loop in which each new genome makes the tool smarter for the next patient. For a disease that kills well over a million people each year, turning sequencing data into faster treatment decisions and earlier outbreak detection is a goal worth the engineering effort, and the updated TB-Profiler brings that goal measurably closer.

Subject of Research: Genomic prediction of antimicrobial resistance and transmission in Mycobacterium tuberculosis using the updated TB-Profiler tool

Article Title: Updated TB-Profiler: enhanced genotypic antimicrobial resistance prediction and relatedness analysis powered by a database of over 171,000 Mycobacterium tuberculosis genomes

Article References: Phelan, J. E., Sawaengdee, W., Thorpe, J., Thawong, N., Piboonsiri, P., Billows, N., Wang, L., van Heusden, P., Meehan, C., Köser, C. U., Mahasirimongkol, S., Campino, S., & Clark, T. G. (2026). Updated TB-Profiler: enhanced genotypic antimicrobial resistance prediction and relatedness analysis powered by a database of over 171,000 Mycobacterium tuberculosis genomes. Genome Medicine. https://doi.org/10.1186/s13073-026-01767-y

Image Credits: AI Generated

DOI: 10.1186/s13073-026-01767-y

Keywords: tuberculosis, TB-Profiler, drug resistance, whole-genome sequencing, Mycobacterium tuberculosis, antimicrobial susceptibility prediction, genomic surveillance, transmission analysis, WHO mutation catalogue, bedaquiline, pretomanid, public health genomics