Imagine a cell. A liver cell or a lung cell, say: a little porous packet of fluid with a tightly packed nucleus, like a baseball's rubber core. Within that nucleus are strands of DNA instructions for life, scrunched together like a balled­up pearl necklace into little ellipsoidal shaped chromosomes.

How DNA is physically arranged in a nucleus­or packed-depends on the cell it is in. Because, as it turns out, nucleus shape varies with cell type, suggesting that the ellipsoids are packed in different ways in different cells. When cells are strained or compressed, however, their nuclei are too. DNA strands get rearranged as a result, changing how they are packed in the nucleus. In turn, those variations in packing put different combinations of genes in proximity to each other, thereby impacting gene expression-even affecting how cancer cells invade and proliferate.

Now, typical human cells have 23 pairs of chromosomes, tens of thousands of genes, and roughly 3.2 billion letters' worth of DNA. How many different ways might mechanical forces affect gene expression when the packing configuration of genetic material shifts?

Exploring questions like these is part of Caroline Uhler's job. Caroline, who is Swiss,joined LIDS in 2015 and is an assistant professor in the Department of Electrical Engineering and Computer Science. Her interest in how gene regulation is affected by chromosome packing is one aspect of her broader interest in graphical models and directed networks.

Gene regulation is an example of a directed network, Caroline explains. In a gene regulatory network, the nodes are the genes, and the edges are the interactions between them, as genes are transcribed to produce proteins that can then up-regulate or down-regulate other genes. Teasing out the structure of the network helps us understand it better.

"When you have thousands of variables, the only way of understanding their relationships is to estimate a network which tells you which nodes are interacting with each other, and that gives you some insight into the complex phenomenon and what is really behind it," she says.

As for the packed-together chromosomes? Caroline models the spatial organization of chromosomes within the nucleus using ellipsoid packing models and predicts how gene regulatory networks change.

Genomics, she notes, is an ideal testing ground for new methods of causal inference for learning directed networks. "I can develop new methodology and make a prediction, and in collaboration with biologists, actually validate the method through interventions such as CRISPR." The gene-editing method CRISPR is able to pinpoint and make changes to individual genes, allowing researchers to study the impact of a single change at a time.

Caroline also works on other types of graphical models-an analytical tool that can be applied just about everywhere. For instance, in weather forecasting you need to know how the weather in one place is correlated with the weather in a different place. You can represent that by a network, too.

She sketches a map of California, harking back to her PhD days at UC Berkeley. "The weather in one city"-San Francisco, say-"is similar to the weather in another nearby city." Her pen trails south to San Jose, and then east, over the Sierra Nevadas. "But the weather on one side of the mountains could be different from the weather on the other side."

Estimating and understanding these networks makes for more accurate weather forecasting, which has practical applications that include mapping out where solar or wind energy are strongest, and where they might be tapped. Caroline currently holds a Doherty Professorship in Ocean Utilization, a two-year grant, which funds a project using data from Austria's meteorological service. "It's been fun to get access to the kind of data they have, and try to improve the weather forecast for a particular region," she says.

Caroline's focus has evolved since her school days, when she was trying to figure out how to combine her interests. "I always loved mathematics. I really liked abstract thinking and I liked science in general, but I was actually also considering studying Latin," she reveals. "What I loved about Latin was exactly what I loved about math-it's like a puzzle where you're trying to figure out how to put together words to build meaning."

Casting about for potential career ideas, she taught for a year between high school and university, and thought she might eventually be a schoolteacher. At the University of Zurich, she studied mathematics and biology, with a particular interest in how statistics tied both fields together. Then, she attended a talk by a visiting professor on Algebraic Statistics in Computational Biology.

"That was everything I loved in one title," she laughs. The speaker, Bernd Sturmfels, eventually became her PhD advisor.

Those varied interests also brought Caroline to LIDS and IDSS (the Institute for Data, Systems and Society), which she feels has a very open, inclusive, and collaborative environment to pursue interdisciplinary work with a solid mathematical foundation.

While Caroline may be settled in at LIDS, she is a seasoned globetrotter. She's lived in seven countries, including Iran, Greece, and Saudi Arabia as a child, as her family followed her father's job with Swissair, and has volunteered in Thailand for half a year. She also has adventurous taste buds and adventurous hobbies, like hiking, kayaking, and generally exploring the outdoors.

Today, at LIDS and IDSS, she hopes to further her work on graphical models, and expand the universe of their applications. It's a fitting task for an up-and-coming scholar with a love of adventure.