About a third of all approved drugs target some kind of G protein–coupled receptor (GPCR). But designing drugs for these receptors poses difficulties because they sit in the cell membrane, and the binding pocket is heavily occluded. Many small molecules for these GPCRs have been identified from nature, but new work uses artificial intelligence design methods to formulate miniproteins that can target GPCRs (Nature 2026, DOI: 10.1038/s41586-026-10656-8).
In the new work, researchers at the Institute for Protein Design (IPD) and AI protein design company Skape Bio, as well as other institutes, were able to develop both antagonists and agonists for a range of GPCRs, including those involved in pain, metabolism, and migraine. It is more complicated to design agonists than antagonists; instead of simply blocking the receptor, agonists must promote a structural change that activates the receptor. These designed agonists accomplish this by stabilizing the receptor’s active form.
GPCRs “are very dynamic molecules—sometimes they assume the active state, even though there’s no molecule there,” says Christoffer Norn, one of the project leaders and CEO of Skape Bio. “So when you then come in with your miniprotein that is made just to fit the active state, you can lock that into place.”
Beyond potential therapeutic targets, binding to these receptors may teach researchers more about how they work. “Many GPCRs remain underexplored because we do not have ligands or tools to explore their function,” says Edin Muratspahić, a researcher at the IPD. “With these computational methods, now we can explore other GPCRs in more detail.”
In an email to C&EN, Lauren Slosky, who works on GPCRs in the pharmacology department at the University of Minnesota Twin Cities and was not involved with the research, praises its breadth. But she also notes that the paper describing the work focuses on GPCRs with protein ligands and relatively accessible pockets. GPCRs with more-occluded or smaller binding pockets are still challenging, she says.
