Diagnosing tuberculosis with algorithms
We've been able to decipher DNA for decades now. Until recently, however, the information behind this blueprint of life was inaccessible. Now mathematical algorithms unlock this information for us. And that has highly promising applications, such as diagnosing complex tuberculosis infections.
The abstract world of mathematics is far removed from a bed in a hospital in a developing country. Take, for instance, the Church of Scotland Hospital in Tugela Ferry, a small country town in South Africa. The hospital was unknown to the rest of the world until it experienced an outbreak of tuberculosis, a contagious bacterial disease, in 2006. This extensively drug-resistant tuberculosis (XDR-TB) proved very difficult to treat and spelt death for most of the patients in the hospital. In almost no time, more than 300 patients with an XDR-TB infection were reported throughout South Africa. For lack of proper treatment, the majority of these patients died.
Now, more than ten years later, this extensively drug-resistant family of TB bacteria is still responsible for a quarter of all XDR-TB cases in the area round Tugela Ferry. The strain has become an example of the major problem that resistance to TB antibiotics presents in developing countries in Africa. Nearly everyone there carries this bacterial infection, which is easily spread from person to person through the air. If a patient goes to hospital for treatment with antibiotics, the doctor cannot see what form of TB the person has. Treatment is therefore completely effective in only 20 per cent of all cases: the person is often prescribed either too much, or not enough, antibiotics. This inadequate treatment can lead to resistance: TB strains that ‘get used’ to drugs and for which increasingly specific antibiotics are needed. This can lead to an infection that is increasingly difficult to control, such as with MDR-TB, which can easily spread within South Africa and beyond.
The idea is “sequencing for anybody, anywhere”. Combined with our algorithms, this give us a test that tells us it's time to see a doctor.
In Delft, Thomas Abeel is working on a solution to this global problem. Together with partners, such as hospitals in South Africa, he diagnoses cases including complex TB infections, which consist of multiple types of TB bacteria. But instead of a blood test or X-rays, he is using algorithms. ‘Complex infections are therefore more difficult to treat, which increases the risk of contagion’, Abeel explains. ‘One of our findings is that there are many more complex infections than we thought. Each strain of bacteria has its own resistance profile. And a combination of strains can lead to a resistance that we are no longer able to treat.’ Such as the XDR-TB infection that startled South Africa. Since then, 10 to 30 per cent of TB infections are classified as complex.
Using patients’ genetic data, the algorithms that Abeel has come up with can 'pull apart’ such a complex infection at genetic level and derive which kinds of TB the patient has in his body. Abeel: ‘An algorithm is a kind of computer program that is based on a model. Like the principle that a sequence of DNA is a combination of DNA from multiple sources. We can turn that into computer code. That enables us to divide the sequence of tuberculosis DNA and find out what the original TB strains were.’ And that is greatly needed, because the number of patients with complex TB infections continues to rise. Abeel: ‘The biggest problem is that 99 per cent of the infections occur in developing countries. Those are the countries with the greatest resistance to antibiotics, but they are also the countries with limited funds for diagnosis and therapy.’
Thomas Abeel's algorithms reveal resistance patterns in TB strains and create prospects for new treatment options. It takes more than just a couple of weeks to treat someone for TB; it is a long and arduous treatment. Abeel: ‘People often take four kinds of antibiotics for two months, followed by two kinds for another four months. For an MDR-TB infection, the regime lasts two years and is often much more complex. The drugs have serious side effects, such as skin discolouration and disorientation. It’s almost like chemotherapy.’ Many patients often feel much better after a week and therefore stop that arduous treatment too early. And that breeds even more resistance. Abeel: ‘If we know exactly which antibiotics still work for each strain, we can diagnose and treat complex TB patients more precisely.’ That increases the chance of being cured immensely.
There’s no reason to assume that we cannot apply our algorithms in other fields.
Nevertheless, Abeel says we may still have to wait a long time before complex TB patients can be treated individually. One of the reasons lies in the technological challenges. For instance, how do you obtain genetic material in a developing country? Abeel: ‘We have all the components, but it will still take several years of technological development before everything can really all be used together.’ One of the most promising developments in this area is nanopore sequencing: unravelling patients’ DNA with a small device that looks like a USB stick that fits in your laptop computer. The company that is developing this is also working on a DNA pen that extracts DNA directly from blood or saliva. The rear of the extraction pen emits a drop of prepared material that is fully suitable for nanopore sequencing. Abeel: ‘The idea is “sequencing for anybody, anywhere”. Combined with our algorithms, this technology gives us a test that tells us: yes, you have TB; time to see a doctor. Or that a doctor can use it to make a diagnosis and a suitable treatment plan.’
Thomas Abeel studied Computer Science Engineering in Ghent. After graduation, he took a PhD in Plant Biotechnology, followed by a three-year internship at the Broad Institute in Boston, part of the Massachusetts Institute of Technology (MIT). It was there that he started his work on the genetic analysis of tuberculosis. Since 2014, he has been Assistant Professor of Bioinformatics in the Intelligent Systems Department of the Faculty of Electrical Engineering, Mathematics and Computer Science at TU Delft. He still holds his guest appointment at the Broad Institute.
But it doesn't stop there, Abeel thinks. His algorithms can be used in an even broader context, such as by applying them to the food industry. ‘Maybe we want to be able to check the chicken in the supermarket for resistant bacteria? Or food producers might want to check their yoghurt cultures or strains of yeast for contamination? DNA sequences form the basis, so there is no reason whatsoever to assume that these algorithms cannot be used elsewhere. That, too, is diagnostics.’
Text: Koen Scheerders
Photo: Mark Prins