A new test could identify tuberculosis in less than an hour

Tuberculosis is the deadliest infectious disease globally, but quickly detecting it — and distinguishing it from other respiratory infections with similar symptoms — has remained a challenge. A new paper describes a test capable of diagnosing TB in less than an hour. Using machine learning techniques to sift through data from 317 patients, the scientists identified four blood proteins that are unique to tuberculosis to develop their test. The accuracy of the test improved when they added a fifth marker that picked up on specific proteins produced in response to toxins expressed by TB. Overall, the new test was correctly able to identify those with TB 86% of the time, and was correctly able to rule out disease in non-TB infected people 69% of the time, which aligns with the WHO’s requirements for a new TB test.