"Respect for animals is a core value for me"

An assistant professor at EPFL's School of Life Sciences, American Mackenzie Mathis has been recognized for several years in Switzerland and internationally for her work in neuroscience. A pioneer in the integration of artificial intelligence in this field and holder of the Bertarelli Foundation Chair in Integrative Neuroscience at Campus Biotech, she co-developed DeepLabCut, an open-source tool that enables automated and non-invasive monitoring of animal behavior. This method of analysis helps reduce stress in experimental animals and improves the quality of data useful to behavioral science researchers. On Wednesday, December 3, the Swiss Swiss Laboratory Animal (SGV) awarded her its 2025 Prize “for her outstanding contribution to the refinement of animal research” and her commitment to the 3R, a concept that aims to reduce the number of animals used, replace them with other models where possible, and refine experiments to minimize stress or pain.

How did you get into research, and more specifically neuroscience?

As a child I was always fascinated by the natural world. I first planned to go to medical school, but after working in an operating room during college and then experiencing working in a research lab as a technician at Columbia University, I realized how powerful basic research could be. Studying stem-cell derived motor neurons (neurons that carry motor commands) pulled me toward fundamental questions about how the motor system works. During my PhD at Harvard I fully shifted to systems neuroscience, as I believed that understanding circuits and behaviors was the most useful way to contribute to science.

What prompted you to pursue AI so early and create DeepLabCut, a tracking tool that is now widely used?

Accurate behavioral quantification felt like a major bottleneck in my research in order to study how the brain generates movement. It became clear that the rapidly advancing field of deep learning might be able to solve this directly. That led to the development of DeepLabCut, which used transfer learning to achieve accurate, markerless pose tracking with very little human-labeled data. Once we saw how well it generalized across videos, and that it was possible for any user to tailor their own custom solutions, it was clear this could change the field, so I was inspired to be sure it was as user-friendly as possible.

Will AI open even more opportunities for your research?

Absolutely. In my lab we developed a new algorithm called CEBRA, which helps us uncover structure in neural and behavioral data that would otherwise remain hidden. As such AI models become more powerful and interpretable, we can push further into understanding adaptive motor control and potentially enable new neuroprosthetics, such as brain-machine interfaces.

Neuroscience still relies heavily on animals. Are these models still appropriate for understanding the human brain?

Non-human animal models remain essential for studying neural circuits and behavior, but we must acknowledge their limits. This is why tools that improve precision and reduce invasiveness matter. They allow us to collect richer data from fewer animals and improve reproducibility. I see AI as a way to reduce the burden while still making progress on fundamental questions, which ultimately shed light on human biology.

Your work also aims to improve animal welfare. Where does this determination come from?

I grew up around animals and have always cared deeply about them. That background clearly influences how I run my lab. We design experiments that minimize stress and use detailed behavioral tracking and high-density neural recordings so we can get the most out of each mouse we use. Respect for animals is a core value for me.

We follow the highest welfare standards and design experiments so that every animal contributes meaningfully to scientific understanding

Do you have other 3R focused projects you would like to mention?

Many of our tools directly support the 3Rs. Markerless tracking reduces the need for invasive approaches. CEBRA allows far more information to be extracted from each experiment. We also work on open source software that democratizes these methods, so that laboratories can refine experiments and reduce animal use globally.

How do you cope with the death or suffering that can occur in animal research?

It is difficult. I cope by staying grounded in the purpose of the work. Neurological diseases cause immense human suffering, and understanding the underlying biology is necessary for future therapies. Like all research groups at EPFL, we follow the highest welfare standards and design experiments so that every animal contributes meaningfully to scientific understanding.

You are one of the most followed EPFL professors on social media. Was that intentional?

No. I wanted to share science openly, highlight the value of open tools, and build a community. As our methods gained popularity, visibility followed naturally. I see it as a way to broaden access to science rather than as an end in itself.

Has working with animals ever caused problems for you on social media?

There is always some risk of misunderstanding or criticism. I choose to be transparent. Showing how we care for animals and why this work matters usually reduces hostility. Open and transparent communication also helps to build trust.

What is life like in Switzerland for a Californian, and how do you see the research environment at EPFL?

Switzerland offers an excellent quality of science, nature, and quality of life. EPFL and Campus Biotech create a unique ecosystem where neuroscience, engineering, and AI interact closely. It is an ideal place to pursue the type of ambitious interdisciplinary work we strive to do.