Diabetes, a widespread disease, is still incurable in most cases. But with the help of data science, this could soon change. At MUDS, PhD student Karin Hrovatin is researching the properties of the many cells involved in insulin production. In the future, her findings should help to stimulate defective cells in the pancreas to produce the vital hormone again.
As a child in Slovenia, Karin Hrovatin loved drawing and math. She was particularly fond of a primary school exercise book called “Math is a Game” that required coloring in pictures to solve math problems – so fond of it that once she finished the exercises in her copy she started again in her classmate’s workbook.
Today, Hrovatin is a PhD candidate at the Munich School for Data Science (MUDS). But she’s still using color, drawing and visualization to solve big problems. “Most people see data science as computer programming, statistics, and math,” Hrovatin says. “But actually, drawing and plotting visualizations is one of the key aspects of data science.”
Using scatter plots that resemble rainbow-colored clouds or abstract art, Hrovatin is taking on a problem that affects nearly 1 in 10 adults worldwide: Diabetes, a chronic condition which can lead to amputations, blindness, heart attacks and strokes. “Diabetes is a very problematic disease,” says Hrovatin. “And it’s very widespread, which makes it a hot topic to try to solve.”
Better understanding the body's insulin factory
After earning a degree in biotechnology at the University of Ljubljana and studying bioinformatics and diabetes at the University of Edinburgh, Hrovatin realized that she was more interested in numbers than hands-on lab work. But the Slovenian data scientist wanted to keep working on diabetes, one of the most pressing problems in public health today.
A unique interdisciplinary data science program run by the Helmholtz Association enabled her to do both, by applying cutting-edge data science methods to better understand the progression of diabetes at the cellular level. Now a PhD student at the Helmholtz Information and Data Science Academy’s Munich School for Data Science, or MUDS, Hrovatin is working on ways to better understand beta cells, the body’s insulin factories. Located in the pancreas, beta cells respond to rising levels of glucose, also known as blood sugar, in the bloodstream by producing the hormone insulin. That, in turn, signals muscle cells to absorb and store blood sugar to use later.
But when the beta cells break down, they stop producing enough insulin – and the body stops absorbing blood sugar. That, in turn, causes type 2 diabetes, one of the most widespread and fast-growing non-communicable diseases in the world today.
To find a cure, researchers are working to better understand why beta cells stop working. The old assumption was that eating too much sugary food forced the beta cells to work too hard, eventually causing them to wear out.
But researchers now know that beta cells aren’t all the same. They vary from person to person, or even between neighboring cells. Understanding what makes beta cells tick – and, sometimes, stop ticking -- is at the heart of Hrovatin’s work.
While researchers once assumed that worn-out beta cells were beyond repair, new discoveries have shown that some beta cells can be revived – but it’s not yet clear which ones respond to treatment. “Beta cells change during aging, or from stress. We’re looking for the differences between healthy and diseased beta cells, so we can figure out how to regenerate cells or bring back their function,” she says. “If you could convert them back to their healthy state, you could restore beta cell function to the pancreas.”
An "atlas" of insulin-producing beta cells
The lab she’s part of at the Helmholtz Center Munich’s Institute for Computational Biology is focused on “single-cell sequencing” data analysis a technique that enables biologists to directly examine cell traits in diabetic mice. Despite its huge potential, the research is conducted on a microscopic scale: Hrovatin’s collaborators extract pancreas cells from the rodents and split the organ tissue up into single cells, to study their properties and how they respond to treatments and stress.
The beta cells are lined up in a tiny tube and encapsulated in oil droplets, then individually “tagged” with a unique identifier. A given data set might be based on just 10,000 cells, an amount barely visible to the naked eye.
Hrovatin’s ultimate goal is to combine data from many different cell types and experiments to create a sort of “atlas” of beta cells, understanding which ones share properties and why their metabolic function differs. That could guide future researchers to better, more personalized treatments for type 2 diabetes. The Helmholtz Diabetes Center is at the forefront of the research. “There’s a lot of data collected already,” Hrovatin says, “which makes diabetes a good problem for data analysis.”
Using machine learning and data science, Hrovatin plans to analyze what cell traits different disease types and treatments have in common – creating a sort of atlas of cell subtypes. “I could use that information to predict how cells will likely respond to treatment in a living organism,” Hrovatin says, “based on how they respond in cell culture.” That, Hrovatin hopes, will be an important first step towards therapies for people with diabetes.
An important step: making data sets comparable
Before she can make meaningful comparisons of cell types and function, though, she has to make sure data sets match up. In an ideal world, all researchers would use the same methods in their experiments, making it possible to easily compare results to see how cells react to different stresses and treatments.
But this is not an ideal world. When she arrived in Munich this summer, she realized the scale of the problem: “There are different mouse models, different disease types, different lab protocols,” Hrovatin says. “There’s a lot of variation, and that’s a big challenge.”
As a result, she’s spent the first four months of her PhD finding ways to make data sets line up, to make sure cells can be compared in a meaningful way. “It’s important to bring the data sets together, but first you have to make sure you’re analyzing biological effects, not technical effects,” Hrovatin says. “I’m hoping to collaborate with biologists to check the results.” So far, she’s been able to draw on the expertise of diabetes and bioinformatics researchers at the Helmholtz Center Munich.
Interdisciplinary suggestions by the MUDS
The Alpine backdrop of Munich reminds the Slovenian researcher of her native Ljubljana, the capital of Slovenia. She’s enjoyed exploring the city by bicycle – and commuting to her lab, despite the strange and isolating conditions of the coronavirus pandemic. MUDS, meanwhile, provides a growing network of other PhD students with similar interests.
Part of the Helmholtz Information and Data Science Academy, MUDS connects the Helmholtz Center Munich with the Max Planck Institute for Plasma Physics and the German Aerospace Center. Her peers include experts in robotics, plasma physics and biomedicine – all united by a shared focus on applying data science in new ways. Seminars bring the different doctoral students together regularly online, until they can start meeting in person. Says Hrovatin: “It’s great – I can learn from people in other fields and get ideas to apply to biology as well.”
Hrovatin still finds time to indulge her childhood passion for drawing. In her free time she sketches fanciful fashion and haute couture, for example. But her energy these days is poured into a different kind of art: Painting colorful visualizations with data points and cell types, hoping to one day heal people with diabetes.