Clear, understandable and interactive images
With the help of data visualization, researchers can create order out of chaos. “For example, computer graphics can be images generated for a cartoon. I work in the field of visualization, which is related to computer graphics,” Thomas explains, “but a little different.” Within visualization, researchers generate images based on data. “Through visualization, we want the information from this data to be as easy as possible for our human visual system to understand,” Thomas explains.
In other words, data visualization is used to create clear, understandable, and interactive images that display datasets containing a large amount of complex information. Thomas is working on algorithms to make these visuals efficient. “I usually work with data that is high-dimensional. This means that each item in the dataset can be described by many variables or attributes,” explains Thomas. This could, for example, be biological data. A data set of DNA sequence measurements may have thousands of genes describing a single sample or item in such a data set. “We humans have a very hard time imagining what such high-dimensional data looks like,” says Thomas. “That’s why I’m developing methods to visualize this data so that we can actually understand it on a 2D computer screen.” But creating the visualizations is only one aspect, Thomas says. “We combine visualization with interaction. We enable the person viewing the visualization to zoom in on the visualized data, to filter by specific properties, or to combine multiple visual views to get different perspectives on the data.”
Visual platform for millions of cells
“Eight years ago I started as a postdoc at TU Delft,” Thomas reflects. He immediately started a collaboration with the Leiden University Medical Center (LUMC). “At the LUMC they had just received a new machine with which they could determine the properties of cells, a so-called Mass Cytometer,” Thomas explains. This machine allows researchers to use blood or tissue samples from volunteers to extract information from individual cells.
At the time, researchers at the LUMC had collected data on millions of cells in a dataset. “I started working on software that they could use in the hospital to view the data,” Thomas says. “And that became what is now known as the Cytosplore platform. That is an interactive visual analysis system that allows researchers to understand how the immune system works. The goal is to provide a clear picture of the composition of the cells of the immune system, and their properties and functionality.” If you look at the Cytosplore platform, you will see that the cells, the proteins and the cell properties are all represented by different colored dots, of different shapes and sizes, and at different distances from each other. You can also zoom in on the data and discover more about the properties of a cluster of cells or proteins. The patterns within these clusters give researchers many insights into the cells.
“When we first started, researchers had thousands of cells in a data set. Now there are millions of cells in a typical data set. The largest data set we had contained roughly 30 million,” he adds. “And the proteins we can measure are currently about 50 per cell. Technically, the software could process even more.”
2024-02-27 20:45:59
#Creating #order #chaos #data #visualization
