Drug-resistant tuberculosis (TB) can be crossed off the list as a threat soon enough. With a new innovation as the potential product of CRyPTIC, a global project that is run by a team of researchers at the University of Oxford, drug-resistant TB can be diagnosed within minutes instead of months.

“It’s rapid,” says Sarah Hoosdally at the University of Oxford, who is managing the project."We're hoping to extract the DNA directly from the sample.”

The project has aimed for a handheld DNA sequencer - as the endpoint product- that analyses a patient's spit sample so that the precise combination of drugs can be prescribed to treat the illness. This will lead to an increased efficiency in treating the patient as the software that is associated with the product will be able to prescribe the right medication just by analysing the genome of the bacteria - although it may be a few years before the product can be used in clinics globally.

“If you can diagnose someone and know their drug resistance profile in less than a day, you’re going to massively improve treatment,” says Hoosdally.

TB diagnosis methods have not changed in more than a century

Tuberculosis is one of the leading causes of death from infectious diseases - alongside HIV.

According to the World Health Organisation, it intends to end the epidemic by 2030 however the rise of superbugs is interfering with that aim, especially in the last decade whereby drug-resistant bacteria have spread through poorer regions of the world, which are densely populated. In 2014 itself, 9.6 million people were infected with TB and 1.5 million of those, died.

“The way that TB is diagnosed is the same way we were doing it when the disease was identified over 130 years ago,” says Marco Schito at the Critical Path Institute in Arizona. “Often individuals pass away while they’re waiting for their result.”

Resistant bacteria can be eliminated with the right cocktail of drugs however as diagnosis takes a long time, either less effective drugs are prescribed, which can risk increase of infection or more effective drugs are prescribed, risking the generation of superbugs.

The identification of bacteria by analysing them after growing them in the lab and staining them with dyes is at least a month-long process. Afterwards, testing them with different combinations of drugs can take up another month.

Faster, more accurate diagnosis for TB

Thus, CRyPTIC aims to address this need for a quicker and more efficient way to work out the exact concoction of drugs required.

Collaborating with laboratories in places where TB is rampant - such as the National Institute for Communicable Disease in Johannesburg, the Chinese Centre for Disease Control and Prevention in Beijing and the Foundation for Medical Research in Mumbai - the team is collecting data on the different types of TB genomes and the specific drugs that each mutation responds to.

The teams send TB genome sequencing results to a machine learning system at Oxford that is being taught about what drugs work on which strains of TB to cut out the process of testing cultures in the lab. It also helps the team at Oxford to untangle the complexity of TB resistance such as two bacterial samples with very slight differences that may resist the same drugs without knowing which genes are involved.

The machine learning takes out the process of guesswork, similar to image recognition software. CRyPTIC trains its artificial intelligence (AI) to recognise drug resistance by supplying it with huge numbers of genomes that are labelled as resistance to a specific drug. The software will afterwards be automatically tuned to recognise the different TB genomes and thus recommend the appropriate drugs.

“The key is getting that catalogue,” says Zamin Iqbal, who works on CRyPTIC’s database in Oxford. “The cherry on top is observation.”

Naturally, the more data, the better. Right now, the primary aim of the project is to speed up diagnosis, however it may have other features such as an early warning system if new strains of tuberculosis are detected and potentially other infectious diseases.

The AI is not expected to completely eradicate tuberculosis as it would require radical social change to address the socio-economic conditions behind the infection, says Nerges Mistry, director of the Foundation for Medical Research in Mumbai.

“It’s a firefighting tech at the moment,” says Mistry. “But we may bring it down, and I think that’s the right thing to do.” MIMS


Read more:
Singaporean with TB who fled CDC gets six month jail term
Malaysian student finds antibiotic alternative in the battle against superbugs
First new antibiotic in a decade aims to counter growing bacterial resistance