A new tool uncovers how a disease develops in individual patients
UiB and Imperial College researchers have developed a tool which predicts how progressive diseases like cancer and malaria develop in individual patients. In addition, the tool uncovers how bacteria develop resistance to certain drugs.
Main content
Mathematicians at the University of Bergen (UiB) and Imperial College, London have developed a brand new tool, which processes data from patients in the thousands into a "roadmap", revealing how diseases develop over time in individual patients. The research is published in the scientific journal Cell Systems.
"Diseases that worsen over time develop very differently from patient to patient. This is one of the great challenges concerning these diseases. The symptoms themselves can be different, appearing in different orders, and they interact with each other. That is what makes it difficult to assess each patient's risk, what the next stage of the disease is most likely to be - not to mention what is the best treatment," Iain Johnston says.
Johnston is an Associate Professor at the Department of Mathematics, UiB and has led the research that created the new tool. He describes it as an algorithm capable of teaching itself a myriad of different disease progressions, based on data from a high number of patients.
The tool is called HyperTraPS, and improves the more data it is fed. It can be used to uncover the main structures of how a disease develops, as well as predict the next step the disease is likely to take, using a given set of symptoms.
"Imagine a great river, branching off across an enormous delta. All the patients start off at the same location, but as the disease progresses, they follow different streams down the river. Our tool teaches itself how the river creates branches, and can predict, with quite high accuracy, where you will end up, based on where you are located in the current right now," Johnston explains.
Used in studies on ovaries and malaria
The researchers have applied the tool on data from several thousand patients og ovary cancer. Cancer causes mutations - harmful aberrations- in our cells, making them divide in an uncontrolled way.
These mutations can vary greatly. The analysis have shown that ovary cancer displays distinctly different disease progressions, highly determined by which mutation ocurred first.
"This shows that this tool can be an important first step on the path to a more efficient, individually tailored made cancer treatment," says Johnston.
REVEALING DISEASE PROGRESSION: The "Waterfall" in the foreground shows paths from one stage of a disease to the next, learnt by HyperTraPS using data from a high number of patients. Each dot of the illustration represents different stages of disease, for example a specific set of symptoms or a given set of mutations. The thickness of the lines indicate the probability of moving from one specific stage of disease to the next.
HyperTraPS has also been used in analysing the clinical data from nearly 3000 Gambian children suffering from malaria. Based on relatively simple data on symptoms, estimating the development of the disease presisely was made possible.
The malaria disease kills nearly 400.000 African children each year, and the World Health Organisation's guidelines on diagnosing and treating it are at the time quite general, making them less efficient in discovering individual disease progressions.
"Based om the study we performed in Gambia, one has now developed a mpre detailed and efficient method of diagnosing and treating malaria. An early identification of hard hit patients is now easier, and scarce resources can be used where they are needed the most," Johnston says.
Uncovering the origin of drug resistance
The biggest strength og HyperTraPS is its wide utility, according to the researchers. In practice, it may be used in mapping any type of process concerning changes over time.
"One example is studying how the organsisms responsible for diseases develop. We have seen how tuberculosis bacteria develop a resistance to drugs, which of course is a major problem for health services," Johnston tells.
Based on samples from 1000 Russian tuberculosis patients, the reseachers have mapped which mutations the bacteria have made as well as the order of the mutations. They found that the bacteria to a lesser degree followed "set runs" in their mutations, making many roads to resistance possible. This is obviously one of the reasons why the problem of resistance is so chellenging to solve.
"The more data we can access, the better we will understand the connections between the different mutations. This can mean a lot for treatment, in the long-run. We might discover that a bacteria that first became resistant to drug A, B and C, in most likelyhood will become resistant to drug D as well. In case, we can choose to give the patient drug K or L instead," the researcher says.
In early 2019, Johnston received the prestigious Starting Grant of 16,8 million Norwegian kroner (ca 1 654 550 Eur) from the European Research Council. The grant will go to further research into applying HyperTraPS in medicinal and biological reseach.