Philip Romero: Using AI to Map and Design New Proteins

Philip Romero joined the faculty of Duke University’s Department of Biomedical Engineering on August 1, 2024. Using both computational and experimental tools, Romero explores how to modify and create proteins with useful applications.

Proteins are large molecules capable of performing a variety of chemical and biological functions, with uses spanning medicine, agriculture, industrial chemistry and even environmental engineering. Despite these broad applications, researchers have struggled to decipher the complex relationship between a protein’s tangled structure and its behavior, limiting efforts to design new proteins with targeted functions. But that’s beginning to change thanks to the development of new computational tools.

In his new role as an assistant professor at Duke, Romero and his team will use these tools to model and explore the structure of proteins with the goal of developing new, more effective therapeutic molecules.

One of the lab’s focus areas is the development and modification of proteases, which are molecular scissors capable of cutting up other proteins. Proteases can be engineered to recognize and inactivate unwanted, disease-associated proteins.

For example, angiotensin converting enzyme 2 (ACE2) can cut Angiotensin-II, a small peptide that regulates blood pressure, to produce Angiotensin-I, which acts as a vasodilator and improves blood flow.

“We can use ACE2 to control blood pressure by cleaving Angiotensin-II. But we’re curious if we could also use the properties of ACE2 for cancer,” said Romero. “If we can localize it to a tumor, we could improve vasodilation, which could enlarge the blood vessels and theoretically make it easier for drugs to reach the cancer cells.”

Romero and his team have also been exploring how they can use AI systems to automate the process of biological engineering. In work published in the inaugural issue of Nature Chemical Engineering, Romero and his team described how they combine AI and a fully automated robotic platform engineer proteins without human intervention. 

“The system is designed to mimic the human scientific discovery process through automated hypothesis generation, hands-on experimentation, and learning from data.,” said Romero. “The AI platform designs proteins and then sends the designed proteins to the robotic system for experimental validation, which then sends the data back to the AI system to update its knowledge and repeat the process.” The system can run continuously days, nights, weekends and can engineer proteins at least 10 times faster than a human researcher.

Prior to his arrival to Duke, Romero was an associate professor of biochemistry and chemical and biological engineering at the University of Wisconsin-Madison. While he recognizes his previous position isn’t traditionally categorized in biomedical engineering, he’s excited about the new opportunities available to him through Duke BME. 

“I earned my bachelor’s degree in biomedical engineering, so this position is like returning to my roots,” he said. “I’m excited to work with colleagues that are interested in computational modeling and more quantitative aspects of biology, and I’m thrilled to be in a department with such a strong connection to medicine and a world-class hospital next door.”