Scientists have devised a blood test for tuberculosis which they hope will help patients in the developing world.
Existing TB tests do not always detect bacteria
The team from St George's Hospital and the Medical Research Council National Institute for Medical Research says its serum test is 94% accurate.
It looks for biomarkers - indicators of infection - and could out-perform alternatives, a Lancet study said.
But a TB expert said many teams had tried to devise such a test but ultimately found they were ineffective.
There are 7,000 new cases of TB reported in the UK each year.
The current method for diagnosing TB involves examining sputum from the lungs under a microscope.
But outside clinics, in rural areas in developing countries, the test is only 40% to 60% accurate.
The researchers found a molecular "signature" for TB.
It was detected using a mass spectrometer and statistical methods to uncover complex patterns known as biomarkers within samples from people infected with TB.
They then used this signature to work out a simpler way to diagnose TB infection.
Dr Delmiro Fernandez-Reyes, of Medical Research Council National Institute for Medical Research, said: "Current developments in mass spectrometry together with new statistical methods for pattern discovery are allowing us to identify biomarkers of infection.
"Our work opens the door to further research into translating these findings to fast and reliable methods of detecting active TB infection that could have a major impact in global health.
"We are currently apply similar approaches to malaria diagnosis."
Professor Sanjeev Krishna of the Centre for Infection at St George's Hospital, who also worked on the study, said: "We are putting forward a fresh approach to an ancient problem.
"I think it is going to be very exciting to make this work in clinics where a test for TB is desperately needed.
"The next stage will be to use the biomarkers we have identified to develop a cheap, accurate and rapid diagnostic test that can be used easily and quickly out in the field."
But Professor Peter Davis, of the organisation TB Alert, said while developing a blood rather than a bacteria test for TB was a widespread aim, it had been a "graveyard" for many researchers.
"The current test isn't very sensitive, and we treat about a third of cases of pulmonary [lung] TB without confirmation of diagnosis from a bacterial test.
"So we do need a new test.
"But not one of the blood tests devised in the past has ever had any value."