In a laboratory at the University of Washington, researchers have cracked a problem that has long vexed medicine: how to design molecules that reliably switch on or off the receptors that govern nearly every function in the human body. The answer came from an unlikely partnership between computational design and artificial intelligence, resulting in custom-built miniproteins that can activate or block G protein-coupled receptors—or GPCRs—with unprecedented precision.
This matters profoundly because GPCRs sit in the membrane that separates the inside from the outside of every living cell, and they are the gatekeepers of sight, smell, hormone sensing, and the body's response to adrenaline, insulin, and countless medicines. Yet for decades, these receptors have been notoriously difficult to control. The signaling switches that activate them sit in deep, flexible pockets—architectural features that make them elusive targets for drug design.
The study, led by David Baker, director of the UW Medicine Institute for Protein Design and a Howard Hughes Medical Institute Investigator, represents a fundamental shift in how scientists approach this challenge. Working in partnership with Skape Bio, Baker's team demonstrated for the first time that artificial intelligence can design miniproteins—proteins with fewer than 100 amino acids—that slip into these hard-to-access pockets and precisely toggle GPCRs on or off. The findings, published in Nature, showcase a generalized approach applicable across different receptor types.
The breakthrough hinges on understanding protein folding in reverse. Rather than studying how proteins fold naturally, the team asked: can AI help us envision entirely new proteins that stick to a target in a purpose-built way? Edin Muratspahić, a postdoctoral researcher and first author on the study, described the moment the team confirmed that their computationally designed miniproteins didn't just bind to these receptors—they actually controlled their signaling in living cells. "Seeing computationally designed miniproteins not only bind but actually control GPCR signaling in living cells was a defining moment for me," he said.
The practical advantages are substantial. Existing drugs like antibodies often fail to activate or block GPCR signaling effectively. In mouse studies, the designed miniproteins performed comparably to clinically used drugs while demonstrating fewer side effects. The team also developed an innovative screening system that tests tens of thousands of candidate proteins directly in living human cells, rather than using purified or artificially stabilized versions that can lose their signaling properties.
This methodological leap matters because it preserves the receptors in their natural state on the cell membrane, where they actually work. Christoffer Norn, corresponding author and co-founder of Skape Bio, emphasized the broader significance: the methods provide "a roadmap for achieving all-computational design of protein ligands for any GPCR."
What began in academia is already moving toward patients. The partnership between the Institute for Protein Design and Skape Bio exemplifies how university research can translate rapidly into therapeutic development. The company is now working to mature these approaches into actual medicines for patients with diseases where GPCRs are effective targets but treatments haven't previously existed. For conditions affecting vision, smell, hormone regulation, and countless other physiological processes, AI-designed miniproteins represent a tangible new frontier in precision medicine.