AI news ·
KAIST and Nobel laureate David Baker use AI to design cortisol-detecting protein and file biosensor patent
KAIST researchers used AI to design a protein from scratch that detects cortisol, the stress hormone, and filed a U.S. patent for a biosensor based on it. The work, published in Nature Communications, was done with Nobel laureate David Baker.
KAIST team designs protein to detect stress hormone using AI
Researchers at KAIST, working with Nobel laureate David Baker of the University of Washington, have used artificial intelligence to design and build a protein that recognizes cortisol, the body's stress hormone. The team has filed a provisional patent for a biosensor based on the protein.
The work, published in Nature Communications, demonstrates a shift in how proteins are created. Rather than searching for natural proteins and modifying them, the researchers used AI to design a protein from scratch with a specific function.
The challenge and the approach
Designing proteins that bind to specific chemical compounds has been difficult because it requires precise calculations at the atomic level. The KAIST team developed an AI model that accurately simulates how proteins interact with ligands-molecules that bind to other molecules.
Using this model, they designed artificial binding proteins for six types of compounds, including metabolites and small-molecule drugs, then tested whether the proteins actually worked as intended.
From lab to application
The cortisol-binding protein became the basis for a biosensor with potential uses in disease diagnosis, drug development, and environmental monitoring. The team filed for U.S. patent protection on the biosensor.
Professor Kyuri Lee, who led the research at KAIST, said the work "experimentally proves that AI can be used to design proteins that precisely recognize specific compounds." She joined KAIST in February 2025 after six years as a postdoctoral researcher and scientist in Baker's lab.
Lee is now directing research on AI-based protein design and artificial enzyme design through KAIST's AI-CRED Innovative New Drug Research Group, funded by the Ministry of Science and ICT's InnoCORE program.
For researchers working on AI for Science & Research, this work illustrates how machine learning models can accelerate the design phase of protein engineering, reducing the need for trial-and-error modification of natural proteins.