Around five years ago, the release of the artificial intelligence system AlphaFold2 drastically changed how protein structures were discovered. In recent months, Anthropic has continued to explore the frontier of protein discovery, using large language models to examine protein binding. While researchers can now model protein structures with AI, they still have a ways to go on predicting how proteins will clump together in diseases, according to Professor of Medical Science Nicolas Fawzi.
A $28.6 million federally funded effort across multiple institutions seeks to address the challenge of predicting protein dysfunction — or the destructuring of proteins and aggregation — in the earliest stages of neurodegenerative diseases such as ALS and Alzheimer’s disease.
Brown’s researchers led by Fawzi will receive a $1 million share of the grant, and the overall research team will be directed by researchers at the University of California, Berkeley.
Fawzi’s team is building a model to identify the problems with proteins that cause specific diseases. The team selected proteins involved in several major neurodegenerative diseases, with a focus on intrinsically disordered proteins, he added. These proteins lack a fixed structure and begin to clump together in some neurodegenerative diseases.
“Instead of staying unbound, they start clumping up and binding to other copies of themselves,” Fawzi said. His lab looks at proteins “one atom at a time.” Researchers are then able to generate data to feed into an AI model, he said.
The team ultimately strives to make a machine learning model that can predict the behavior of these intrinsically disordered programs, according to Fawzi.
Collaborators at other institutions will run computer simulations of the same proteins, Fawzi added. Together, the teams will feed that data to train new machine learning models.
In the project’s third and fourth years, the team hopes to be able to predict interventions — new therapies and drugs — Fawzi said.
Fawzi is well-known for his work studying protein aggregation in neurodegenerative disorders, according to Bess Frost — the director of Brown’s Center for Alzheimer’s Disease Research and a professor of molecular biology, cell biology and biochemistry. Frost, who is not part of Fawzi’s lab, added that Fawzi had been on the Clarivate highly-cited researchers list for two years in a row now.
Katelyn Newton GS, a second-year Ph.D. student in the Therapeutic Sciences Graduate Program, studies the protein TDP-43 in Fawzi’s lab. The protein is vital for healthy cells, but if it misfolds, it aggregates in diseases such as ALS.
Newton’s work hopes to understand how this aggregation forms and causes disease in order to engineer the protein to be “less aggregation-prone,” she wrote in an email to The Herald.
To begin developing treatments for neurodegenerative diseases, scientists must understand how these proteins behave, Frost said. “We can’t make drugs that stop these proteins from aggregating if we don’t understand how they aggregate in the first place,” she said.
Newton added that one of the Fawzi lab’s strengths in the face of complex scientific research is the culture of support among the lab members.
“The types of proteins that we study can be difficult to work with, but we have a lot of camaraderie in the lab and people are always willing to help each other,” Newton said. “Dr. Fawzi is a really knowledgeable mentor and his excitement about science is evident through the endless ideas for experiments he has.”
Elizabeth Rosenbaum is a senior staff writer covering science and research.
