Chase Freschlin, an IPiB graduate student, will be defending her Ph.D. research on April 8, 2025. Her research in the Romero Lab uses machine learning to map relationships between a protein’s sequence and its functions in a cell, which could be used to inform drug development, lab experiments, and other industrial and research applications.
Freschlin worked with a small binding protein, GB1, that is just 56 amino acids long. “Even with a small protein, there are so many possible unique sequences that it’s impossible to study them all. Machine learning helps by identifying potentially functional sequences,” explains Freschlin.
Using GB1 proteins with one or two mutations, Freschlin was able to extrapolate which of GB1’s 1081 possible mutations may optimize a function of interest — and, just as importantly, which may be detrimental. Different machine learning models may learn different aspects of a protein’s function. Freschlin used different models to identify both functional optimization (in the case of GB1, a strong affinity for immunoglobulin binding) and structural stability. Her research has been published in Nature Communications.
When identifying potentially beneficial mutations using machine learning, Freschlin says, the next step is to test the function of mutant proteins. This requires using customized DNA to produce the proteins of interest, which can be expensive. Freschlin developed a gene assembly method using Golden Gate cloning called OMEGA. The new method yields a 10-fold reduction in synthesis costs.
“I was working on a different project engineering proteases, but they were too big and expensive to synthesize,” says Freschlin. “Instead of limiting myself to focus on a specific region to reduce costs, I decided to work on developing a new gene assembly method so that more scientists can do the work of evaluating the sequences designed through machine learning.”
Looking towards future research, Freschlin says, “Different machine learning models may capture different aspects of protein function like binding or stability. In the future, combining models could provide a more comprehensive picture of protein function.”
After graduating, Freschlin plans to continue on to a postdoctoral researcher position and later a career in academia.
Freschlin joined the IPiB program at the height of the COVID-19 pandemic. In search of community and opportunities to gather with other graduate students, Freschlin enjoyed taking long bike rides with fellow IPiB members. When the Romero Lab moved to Duke University in 2024, Freschlin chose to stay in Madison where she had strong connections with her cohort and remaining labmates.
To learn more about Freschlin’s research, attend her Ph.D. defense, “Data-driven protein engineering: computational and experimental methods to explore protein sequence space” on Tuesday, April 8 at 10:00 a.m. CT in Room 1211 of Hector F. DeLuca Biochemical Sciences Building.
Written by Renata Solan.