AI-Powered Breakthrough Enables Drug Design Against Previously "Undruggable" Disease Proteins

A groundbreaking study published in Nature Biotechnology demonstrates how artificial intelligence is being reimagined to tackle some of medicine’s most persistent challenges. Researchers have developed an AI tool called PepMLM that designs small, drug-like molecules capable of binding to and breaking down harmful proteins—even when the proteins’ 3D structures are unknown or unstable. The innovation could unlock new treatments for diseases long considered “undruggable,” including certain cancers, neurodegenerative disorders like Huntington’s disease, reproductive conditions, and viral infections. The work was led by a multi-institutional team from McMaster University, Duke University, and Cornell University. PepMLM builds on language-processing algorithms used in chatbots and other AI systems, but instead of understanding human language, it learns the “language” of proteins—interpreting amino acid sequences to predict how peptides (short chains of amino acids) will interact with target proteins. This approach marks a significant shift from traditional drug discovery, which relies heavily on knowing a protein’s 3D structure. In 2024, the Nobel Prize in Chemistry was awarded to Google DeepMind researchers for AlphaFold, an AI that predicts protein structures with remarkable accuracy. However, many disease-relevant proteins—especially those involved in cancer and brain disorders—lack stable or well-defined shapes, making them difficult to target with conventional methods. PepMLM overcomes this limitation by working directly with protein sequences, bypassing the need for structural data. “Most drug design tools depend on knowing a protein’s 3D shape, but many of the most important disease targets don’t have stable structures,” said Pranam Chatterjee, senior author of the study and now a faculty member at the University of Pennsylvania. “PepMLM changes the game by designing peptide binders using only the amino acid sequence.” In laboratory experiments, the team successfully designed peptides that bind to and degrade disease-associated proteins. These included proteins linked to Huntington’s disease, cancer, reproductive disorders, and live viruses. The results suggest the tool can be used to neutralize toxic proteins, stabilize beneficial ones, or modulate their activity—offering versatile therapeutic potential. Christina Peng, a Ph.D. student in the Truant Lab at McMaster University, led the Huntington’s disease experiments. “It’s exciting to see how these AI-designed peptides can actually function inside cells to break down toxic proteins,” she said. “This could be a game-changer for diseases where traditional drugs have failed.” At Cornell, Matthew DeLisa and Hector Aguilar’s labs engineered and tested the peptides against viral proteins, while Duke’s team developed the AI model and conducted initial validation. Ray Truant, a professor at McMaster, emphasized the broader implications: “This work shows we can now bind any protein to any other protein. We can degrade harmful ones, stabilize helpful ones, or control their modifications—depending on the treatment goal.” The researchers are already advancing the technology with next-generation AI models, including PepTune and MOG-DFM, aimed at improving peptide stability, targeting accuracy, and delivery within the body. “Our ultimate goal is a general-purpose, programmable peptide therapeutic platform—one that starts with a protein sequence and ends with a real-world drug,” said Chatterjee. The breakthrough represents a major leap toward making personalized, precision medicine more accessible and effective for a wide range of conditions.