Johns Hopkins University is leading a four-year research initiative backed by up to $52.1 million from the Advanced Research Projects Agency for Health (ARPA-H) to better understand how proteins change shape and contribute to disease. The project will use artificial intelligence to study protein misfolding, with the goal of developing earlier diagnostic tools and more targeted treatments for conditions including Parkinson’s disease and Ewing sarcoma, a rare cancer affecting bones and soft tissue.
A Multi-Institutional Effort to Decode Protein Behavior
Led by Xiaobo Mao, Ph.D., professor of neurology at the Johns Hopkins University School of Medicine, the initiative brings together 21 investigators across 12 institutions in the United States and Europe.
The project, known as Dynamic Ensemble-to-Rescue: Decoding and Controlling the IDR Molecular Grammar at Scale (DETR), is part of ARPA-H’s broader BIOmolecular Grammar for protein Aggregation Modulation and Intervention (BIOGAMI) program.
The program focuses on understanding intrinsically disordered proteins, which can adopt different shapes depending on their environment. When these proteins misfold or accumulate abnormally, they can contribute to diseases such as Parkinson’s and Alzheimer’s.
Unlike traditional approaches that primarily examine a protein’s sequence, the DETR team aims to predict how these proteins behave under different conditions and identify when they may become harmful.
Using AI to Support Earlier Diagnosis and Treatment
The research will initially focus on Parkinson’s disease and Ewing sarcoma, with the goal of developing technologies that could eventually support other diseases associated with protein misfolding.
The team plans to develop a cerebrospinal fluid test capable of detecting harmful forms of alpha-synuclein, a protein associated with Parkinson’s disease. Researchers hope the approach could eventually identify the disease years before symptoms appear.
The project will also explore molecules that selectively target harmful protein structures while preserving healthy forms. An open-access AI foundation model is planned to help other laboratories investigate disease-related proteins and identify potential therapeutic targets.
“If we can predict the shape, we can design a drug that recognizes the harmful form and leaves the healthy one alone.”
The proposed diagnostic and therapeutic tools remain research goals rather than validated clinical products.
Building a Foundation for Future Therapies
The ARPA-H-funded project will proceed through two 24-month phases. The first will focus on understanding how protein sequences, structural characteristics and environmental conditions influence protein behavior. Researchers will combine laboratory experiments with computational modeling to build and evaluate an AI foundation model.
The second phase will focus on applying that knowledge to detect harmful protein folding and develop approaches that could prevent or reverse it.
The team includes researchers from Columbia University, MIT, Emory University, the University of Michigan and several other institutions in the United States and Europe. Johns Hopkins Medicine will serve as the lead institution coordinating the research.
