Scientists at the University of Virginia School of Medicine have developed a bold new approach to drug development and discovery that has the potential to dramatically accelerate the development of new drugs.
UVA’s Dr. Nikolay V. Dokholyan and colleagues have developed a suite of artificial intelligence-powered tools called YuelDesign, YuelPocket, and YuelBond. Together, these tools will transform the way new drugs are manufactured. At its core, YuelDesign uses a cutting-edge form of AI called diffusion modeling to design new drug molecules tailored to precisely fit protein targets, taking into account how proteins bend and change shape during binding.
A companion tool, YuelPocket, pinpoints where drugs can bind on proteins, while YuelBond ensures that the chemical bonds in the designed molecules are accurate. This combined approach is expected to improve both how new drugs are designed and how existing drugs can be quickly and efficiently evaluated for new purposes.
Think of it like this: Other methods try to design a lock key that is completely stationary, but inside the body the lock is constantly shaking and changing shape. Our AI designs the key as the lock moves, making the fit more realistic. This could be a game-changer for patients with cancer, neurological disorders, and many other conditions. Despite the dire need for better drugs that target these unstable proteins, the impasse continues. ”
Dr. Nikolai V. Doholyan, UVA Department of Neurology
Pitfalls in drug development
The average cost of developing a new drug is estimated to be more than $2.6 billion, and nearly 90% of new drugs fail once they reach human testing. This is due in no small part to the difficulty in predicting how the molecules within a drug will interact or bind to their targets in the body. If a molecule does not bind in exactly the right place as intended, the drug may not work or may have unwanted and harmful side effects.
Artificial intelligence has helped solve this problem and greatly accelerated drug design, but Dokholyan’s research takes it to the next level. His YuelDesign overcomes the limitations of existing options by designing drug molecules while treating proteins as flexible, dynamic structures rather than the rigid, frozen snapshots used in other methods. This is very important because proteins often change shape when a drug binds to them, a phenomenon known as “induced fit.” If you ignore this flexibility, what looks promising on your computer screen can fail in real life.
Dokholyan and his team designed YuelDesign specifically to overcome this problem. The technology uses an advanced AI “diffusion model” to simultaneously generate both the protein pocket structure and the small molecule that can be inserted into it (the key that turns the lock), allowing both to be adapted to each other during the design process.
A companion tool, YuelPocket, uses graph neural networks to pinpoint where drugs should bind on proteins, even in protein structures predicted from existing tools such as AlphaFold. Researcher Dr. Jian Wang said, “Most existing AI tools treat proteins as frozen images, but biology does not. Our approach allows proteins and drug candidates to evolve together during the design process, just as they occur in the body.” “For example, when designing a molecule for a well-known cancer-related protein called CDK2, we found that only YuelDesign could capture the important structural changes that occur when the drug binds.”
In a new scientific paper outlining the YuelPocket test, researchers note that mapping protein pockets is important for “nearly every aspect of modern development.” With these promising results, Doholyan expects this technology to reduce drug development costs, improve the success rate of new drug candidates, and shorten the time it takes for new treatments and treatments to reach patients. (Accelerating how quickly lab discoveries can be turned into medicines to benefit patients is the primary mission of UVA’s new Paul & Diane Manning Institute for Biotechnology.)
“Our ultimate goal is to make drug discovery faster, cheaper, and more likely to be successful, bringing promising treatments to patients faster,” Doholyan said, adding that he wants to “democratize” drug discovery by giving scientists new tools at their fingertips. “We’ve made all of our tools freely available to the scientific community. We want researchers around the world to be able to use them to tackle the diseases that matter most to patients.”
Publication of survey results
Dokholyan and his team have described the development and results of these tools in papers in the scientific journals PNAS, JCIM, and Science Advances. The research team includes Wang, Dong Yan Zhang, Shreshty Budakoti and Dokholyan. The scientists have no financial interest in this research.
This research was supported by National Institutes of Health grant 1R35 GM134864. National Science Foundation, Grant 2210963. Huck Institute for Life Sciences. and the Passant Foundation.
sauce:
University of Virginia Health System
Reference magazines:
Wang, J., Dohoryan, N.V. (2026). An integrated protein-small molecule graph neural network for binding site prediction. Proceedings of the National Academy of Sciences. DOI: 10.1073/pnas.2524913123. https://www.pnas.org/doi/10.1073/pnas.2524913123
