AI Breakthrough Designs Peptide Drugs Targeting Previously Untreated Proteins

Research published in Nature Biotechnology It uncovers powerful new uses of artificial intelligence: even when scientists don't know what those proteins look like, they design small drug-like molecules that can stick to and break down harmful proteins in their bodies. Breakthroughs could lead to new treatments for diseases that have long resisted traditional drug development, including certain cancers, brain damage and viral infections.

This study was published on August 13, 2025 by a multicenter team of researchers from McMaster University, Duke University and Cornell University. An AI tool called PEPMLM is originally based on an algorithm built to understand human language and be used in chatbots, but was trained to understand the “language” of proteins.

In 2024, the Nobel Prize in Chemistry was awarded to researchers at Google Deepmind to develop Alphafold, an AI system that predicts the 3D structure of proteins. However, many disease-related proteins, including those involved in cancer and neurodegeneration, do not have a stable structure. That's where PEPMLM is where you take a different approach. Instead of relying on structure, the tool designs peptide drugs using only protein sequences. This allows for targeting a much wider range of disease proteins, including those previously considered “continuous.”

“Most drug design tools rely on knowing the 3D structure of proteins, but many of the most important disease targets do not have a stable structure,” said Pranam Chatterjee, a senior author of the study who led the study at Duke and is now a faculty member at the University of Pennsylvania. “PEPMLM changes the game by designing peptide binders using only the amino acid sequence of the protein,” says Chatterjee.

In lab testing, the team showed that PEPMLM can design peptides (short chains of amino acids) that help to stick to disease-related proteins and in some cases destroy them. These included proteins involved in cancer, reproductive disorders, Huntington's disease, and even live viral infections.

This is one of the first tools that allows you to design these types of molecules directly from the sequence of proteins. It opens the door to faster and more effective ways to develop new treatments. ”

Pranam Chatterjee, senior author of the study

This study included key contributions from McMaster University, where Christina Peng, a doctoral student at Truant Lab, led the Huntington disease experiment.

“It's exciting to see how these AI-designed peptides actually work intracellularly to break down toxic proteins,” Peng said. “This could be a powerful new approach to diseases like Huntington, where traditional drugs were not effective.”

Other parts of the study were conducted at Cornell. At Cornell, the labs of Matthew Delisa and Hector Aguilar constructed and tested peptides on viral proteins. This study also included contributions from McMaster's Ray Truant.

“This study shows that proteins can now be bound to other proteins,” said Truant, a professor in the Department of Biochemical Sciences. “It can break down harmful proteins, stabilize beneficial proteins, and control protein modifications, depending on your treatment goal.”

The team is already working on next-generation AI algorithms such as Peptune and Mog-DFM, which improves how these peptides behave in the body, making them more stable, targeted and easier.

“Our ultimate goal is a general purpose, programmable peptide therapy platform, which starts with a sequence and ends with an actual drug,” Chattersee said.

The research was supported by the CHDI Foundation, the Wallace H. Coulter Foundation, the Heartwell Foundation, the National Institutes of Health, and the Clenville Foundation in Toronto. Chatterjee and First Author Tianlai Chen are co-inventors of US patent applications related to PEPMLM. Chatterjee and co-author Delisa are Ubiquitx, Inc., a biotechnology company that develops programmable protein-based therapies. brings economic benefits to the