A new $95 million research initiative in Seattle will use artificial intelligence to design biological molecules that do not exist in nature. The AI BioDesign project, announced Thursday and funded for at least five years, pairs AI with bioscience to pursue applications ranging from cancer treatments to enzymes that break down plastics.
The Fund for Science and Technology, a nonprofit foundation launched by the estate of late Microsoft co-founder Paul Allen, is backing the effort. The Allen Institute will receive $46.1 million, the University of Washington will get $43.8 million, and the Fred Hutch Cancer Center will receive $4.7 million.
A closed-loop design process
AI models will propose new biological designs that scientists then build and test in the lab. Results from those experiments feed back into the models, sharpening their ability to make better designs in subsequent rounds. The core aim is to explore possibilities that evolution has not produced and, along the way, learn more about the fundamental rules that govern how biology works.
Evolution creates diversity, but "it is a slow process," said UW researcher David Baker, who won the 2024 Nobel Prize in Chemistry for pioneering the computational creation of new proteins. The project accelerates that exploration through iterative machine learning. For scientists and researchers working at the intersection of computation and biology, this marks a significant expansion of AI for Science & Research.
First targets and practical goals
Jay Shendure, lead scientific director of the project, outlined several early targets. These include custom proteins designed to latch onto disease, genetic switches capable of turning genes on or off, and tools that can selectively destroy or stabilize proteins. Each target represents a building block that could feed into medicines, materials, or environmental technologies.
The work builds on techniques that research scientists increasingly need to understand as AI becomes embedded in laboratory workflows. An AI Learning Path for Research Scientists can help professionals bridge the gap between computational modeling and experimental validation that projects like this demand.
Why this matters for scientists and researchers
The AI BioDesign project tests whether AI can systematically generate functional biological designs that outperform what nature has produced. For researchers, the practical takeaway is clear: the feedback loop between computational prediction and wet-lab testing is shortening. Teams that can operate across both domains will be positioned to move faster on drug targets, industrial enzymes, and gene-editing tools. The project's five-year timeline also means published methods and open-source models will likely filter into the broader research community long before the work concludes.
Your membership also unlocks:
