Infectious
Tuberculosis
A chronic, recurrent infection caused by the bacterium Mycobacterium tuberculosis. Tuberculosis (TB) may affect almost any tissue or organ of the body with the lungs being the most common site of infection. The clinical stages of TB are primary or initial infection, latent or dormant infection, and recrudescent or adult-type TB. Ninety to 95% of primary TB infections may go unrecognized. Histopathologically, tissue lesions consist of granulomas which usually undergo central caseation necrosis. Local symptoms of TB vary according to the part affected; acute symptoms include hectic fever, sweats, and emaciation; serious complications include granulomatous erosion of pulmonary bronchi associated with hemoptysis. If untreated, progressive TB may be associated with a high degree of mortality. This infection is frequently observed in immunocompromised individuals with AIDS or a history of illicit IV drug use.
Living starter kit
Not yet expert-checkedUpdated Nov 13, 2026Start building on tuberculosis
Tuberculosis has some of the best AI-ready data of any infectious disease: bacterial genomes paired with lab-measured drug resistance, linked patient records and images, and hundreds of thousands of cough recordings. The hard problems are screening tools that work in clinics they weren't trained on, and reading drug resistance from a genome fast enough to guide treatment.
Know tuberculosis well? No expert has checked this kit yet.Check our work →Open datasets4
- CRyPTIC drug-resistance genomes ↗
Genomes of TB bacteria paired with lab-measured drug resistance for each sample, from the CRyPTIC Consortium, in versioned downloads.
GenomicsOpen downloadCC-BY 4.0
- NIAID TB Portals patient data ↗
De-identified TB patient records linked to chest images and bacterial genomes, from the U.S. National Institute of Allergy and Infectious Diseases; you request access by describing your project.
ClinicalApply for accessLicense not stated
- CODA TB cough recordings ↗
Cough recordings from TB clinics, labeled with lab-confirmed TB status, from the CODA TB DREAM challenge; access is by request now that the challenge has ended.
733,756 cough sounds from 2,143 patients across 7 countries
AudioApply for accessLicense not stated
- WHO global TB database ↗
Country-by-country TB case reports and burden estimates from the World Health Organization, as downloadable CSV files.
EpidemiologyOpen downloadLicense not stated
Benchmarks and tasks2
- CODA TB DREAM Challenge ↗Predict whether a patient has TB from recordings of their cough, with or without basic clinical details. Audio
- NLM Montgomery and Shenzhen chest X-ray baselines ↗Classify chest X-rays as showing TB or not, and outline the lungs, comparing against the published baselines for the NLM Montgomery and Shenzhen image sets. Imaging
Open questions4
- 01
Can a cough recording screen for TB reliably across countries, phones and clinics?
A cheap, quick screen could find people to test where there's no X-ray machine, but earlier studies were small and limited to a few sites.
AudioClinicalSource: CODA TB DREAM Challenge ↗
- 02
Can a model predict how resistant TB bacteria are to each drug, not just whether they're resistant, from the genome?
The level of resistance, not just yes or no, helps choose drugs and doses, and CRyPTIC is one of the few datasets that measures it.
GenomicsSource: CRyPTIC datasets ↗
- 03
Can combining X-rays, genomes and clinical records predict who will do poorly on treatment?
TB Portals links all three, but its cases were picked for study rather than sampled at random, so models must avoid learning that selection.
ImagingGenomicsClinicalSource: TB Portals data collection ↗
- 04
Can a model tell a real change in TB cases from late or incomplete reporting?
WHO's recent counts are provisional and get revised, which makes it hard to spot disruptions to TB services early.
EpidemiologySource: WHO tuberculosis data ↗
Who's working on this4
CRyPTIC Consortium
An international consortium that shares TB genomes paired with lab-measured drug resistance.
Its versioned data tables are the most direct starting point for drug-resistance models.
No public invitation foundEvidence, Sep 13, 2026 ↗
NIAID TB Portals Program
A NIAID program that collects de-identified TB patient records, chest images and bacterial genomes.
One of the few places where clinical, imaging and genomic data come linked for the same patients.
No public invitation foundEvidence, date unknown ↗
CODA TB DREAM collaboration
Global Health Labs, Sage Bionetworks, UCSF and the University of Montreal ran a challenge to detect TB from cough sounds.
Published results and post-challenge data access give audio models a ready-made baseline.
No public invitation foundEvidence, Aug 28, 2024 ↗
WHO Global Programme on Tuberculosis and Lung Health
Maintains the global TB case reports and burden estimates.
Its tables support forecasting and work on incomplete reporting.
No public invitation foundEvidence, date unknown ↗
Not in the kit yet
- Most datasets here need an access request or don't state a license; check each source's terms before commercial use.
- The widely used NLM chest X-ray sets (Montgomery and Shenzhen) appear only as a benchmark, because we couldn't confirm their current download page.
- Popular community X-ray datasets, such as TBX11K, haven't been checked yet.
- We haven't yet found a group building TB tools with an open invitation to collaborate.
This kit keeps growing as new data, benchmarks and teams appear. Know something we missed? Tell us →
Curated by the Unsolved project team. Every link was checked on the date shown; licenses are as stated by each source.
Where the field stands
01
Early understanding
No strong gene target or drug candidate yet
02
Target found
At least one gene target with strong evidence
03
Drug candidates
Preclinical or phase 1 drug candidates
04
In trials
Drug candidates in phase 2 or later
05
Has approved drugs
At least one approved drug; not necessarily a cure
Derived from Open Targets drug and target data, not self-reported.
Healthy years lost (DALYs), 2023
43.6M
95% range 35.3M to 53.6M
Deaths, 2023
1.0M
People living with it, 2023
1.9B
GBD cause: Tuberculosis (exact match)
Research attention
What there is to build on
Open Targets Platform ↗Gene targets with strong evidence
4
of 5,049 associated genes
Top associated genes
NR3C1LTFKPNB1KPNA1TLR2
Drugs and clinical candidates
134
Approved drugs
27
Identifiers
- Open Targets
- MONDO_0018076
- MeSH
- D014376
- ICD-10
- A15-A19
- Orphanet
- ORPHA:3389
Trial and paper counts are rough measures of attention, not of progress. This page is for research planning, not medical advice.