Of Mice And Machine Learning

Don't dwell on what you don't know, advises Varun Chamarty '26 (ENG).

"Some people, they'll come up with an idea, and then they'll kill it in their own brain," says Chamarty, a pre-med rising junior studying biomedical engineering at UConn.

"I think the best way to begin is by just beginning. You have an idea. You run with it. Trial and error. See what works. See what doesn't work."

That, ultimately, is what all scientific research boils down to: An idea that you test, a hypothesis that you investigate, with the hope of learning, finding, or discovering…something.

And that's what Chamarty has been doing this summer - testing an idea, just to see what happens, with the hope of contributing in some way to an active and growing field of study.

Before he graduated from Farmington High School, Chamarty discovered his own interest in robotics and also took part in Cutting Edge, a program that pairs high school students with graduate students at UConn Health, offering the opportunity to learn about things like biotechnology, bioinformatics, and other DNA-centric technologies.

"Admittedly, it wasn't my thing," Chamarty says, but what did interest him was technology and research - neuroscience, in particular.

"Neuroscience is so different from the rest of anatomy," he says. "You'll take an anatomy class, and if it's two semesters, you'll learn one semester about the whole body. And then, the other semester will be all about the brain, because there's so many things there. And there's even more that we don't know."

And learning things about the brain is exactly what assistant professor Timothy Spellman's lab does at UConn Health.

As a first-year student, Chamarty found and applied to work in the Spellman Lab, which focuses on the physiological substrates of executive functioning within higher-order association areas of the brain. Executive functioning includes complex cognitive functions, like memory, self-control, and flexible thinking.

But a large portion of the Spellman Lab's research is dedicated to another executive function - attention - and figuring what physiologically happens within the brain when someone's attention shifts.

When we're watching something, then listening to something. Switching focus rapidly. Multitasking.

"The reason we're looking at that is because those pathways are poorly understood as of now," Chamarty explains. "So, we're trying to figure out if there are consistent regions of the brain that are activating when doing that.

"And that's interesting on a larger scope, clinically, because people with ADHD or schizophrenia, they have trouble doing that. If we're able to find that pathway, then we can treat it," he says.

In the Spellman Lab, mice perform tasks where their attention is shifted from one stimulus to another - one lickspout or another based on a certain signal, like a whisker stimulation or an odor stimulation - while the activity happening in their brain is monitored in real-time.

For his project, though, Chamarty is bringing a new technology into the lab's mix: machine learning.

A type of artificial intelligence, machine learning focuses on building computer systems that learn from data. Algorithms are trained to find relationships and patterns.

In the case of experiments like those in the Spellman Lab, Chamarty's hope is that machine learning will help researchers make discoveries that they might otherwise not be able to see, using new tools to build on a growing body of neuroscientific research.

While the mice perform their attention-related tasks, two digital cameras capture how they respond to the stimulus, how their attention shifts, and how their physicality changes. Do they twitch their whiskers a certain way? Is there a recognition pattern on their face?

"The reason we're doing this is because we are trying to see if there is a correlation between brain activity and their facial expressions," he says.

Based on years of academic experience and a body of peer-reviewed literature developed not only at UConn Health but also in labs around the world, Spellman and Chamarty suspect some relationships might be there. But they don't yet know if any correlations actually exist or what exactly they might be.

"That's the nice thing about machine learning," he says. "You don't necessarily have to know what you're looking for. You can just see if there are patterns that show up."

He plans to match the machine learning findings against other data developed in the lab to see if any patterns found by the algorithms align with the brain-scan data from other experiments.

While he says that, ideally, he'd one day be able to write and publish about the techniques or the findings of his research, the work he's doing has had a much more immediate impact on him, personally.

He's had to find his own way in the lab, something he says was daunting, at first.

"In high school, there's a 'right answer,' or there's something you were assigned to do," he says. "But, I came in, on my first few days, and Tim was like, 'Here's your project. Good luck.' There's a lot of navigating that you have to learn how to do and, as a freshman, that was kind of new to me."

He didn't know anything about mice, very little about neuroscience, and had never done a project with machine learning.

But his background in engineering - where he's asked to solve problems, often with little or no guidance on how to get there or what the answer might be - was an asset.

"Being afraid of being wrong is something that kills a lot of people's motivation to do new things," Chamarty says. "And I have kind of learned that I don't know anything, so why not? I came in knowing that I wouldn't know what I was doing, but knowing that I was willing to figure it out."

He's still figuring out where he eventually wants to go. He's always thought his path was in medicine - he's also an EMT, so maybe emergency medicine, and he likes radiology.

But in neuroscience, he's also found a way to apply his skills as an engineer.

It's been nice, he says, to have a place like the Spellman Lab where he's been able to explore the worlds of both medicine and engineering as an undergraduate - and where he's been free to learn about a lot more than just the ways that mice twitch their whiskers, how machine learning might be applied in all sorts of research contexts as a tool for discovery, or what it all might mean for future medical breakthroughs.

"I think that I've learned a lot about resiliency," Chamarty says. "You don't always see what's ahead, and that's something you learn in research. I don't really know where it's going, but every day I come in, and I work toward something. Just doing that is exciting, because you don't know what you'll find.

"That's something I want to focus on more - not being so attached to the results, but enjoying the journey to get there and learning from that."

Chamarty has received support for his research through grants from UConn's Health Research Program and Summer Undergraduate Research Fund