AI-Designed Miniproteins Successfully Target Complex Cell Receptors

G protein-coupled receptors (GPCRs) regulate essential physiological processes and are central targets for drug discovery and development. However, designing protein-based drugs for them is notoriously difficult because these receptors are embedded directly in cell membranes and are constantly shifting shapes. Overcoming these hurdles could unlock highly precise therapies for a wide range of diseases. Demonstrating a major leap forward, scientists report today in Nature that they have successfully built "miniproteins" from scratch to bind and control these elusive receptors. The work, led by senior authors Christoffer Norn and David Baker at the University of Washington, with Edin Muratspahić as lead researcher, outlines a new blueprint for creating highly targeted biological drugs.

The research team deployed advanced computational algorithms to design tens of thousands of miniature proteins tailored to fit deep into the binding pockets of various GPCRs. To find the designs that actually worked, they developed a novel, high-throughput microscopy screen that tracks whether a designed protein successfully intercepts its target receptor inside living human cells. This pipeline yielded a diverse array of miniproteins: agonists that activate receptors involved in itch and pain, and antagonists that block receptors tied to cancer, metabolic disorders like diabetes, and migraines. High-resolution cryo-electron microscopy confirmed that the real-world structures of five miniprotein-receptor complexes closely matched their computer-generated models.

In preclinical tests, a custom miniprotein designed to block the CXCR4 receptor effectively mobilized blood stem cells in mice—a process routinely used to collect cells for transplantation in hematological cancers. The designed miniprotein matched the performance of plerixafor (AMD3100), an existing clinical drug, but with fewer adverse effects, avoiding the abnormal spikes in circulating white blood cells seen with the standard treatment.

While this approach proved highly effective for GPCRs that naturally bind peptides, the researchers caution that designing proteins for receptors with narrower, small-molecule pockets will pose a steeper challenge. Additionally, the dynamic nature of GPCRs means that predicting exactly whether a designed protein will act as an activator or a blocker based solely on static receptor structures requires further refinement.

This ability to computationally build functional miniproteins represents a step change for membrane receptor pharmacology. Because these molecules are smaller, more stable, and highly specific, they offer a promising new path for developing safer and more effective therapeutics.

Reference:
Muratspahić, E., Feldman, D., Kim, D.E. et al. De novo design of miniproteins targeting GPCRs. Nature (2026). https://doi.org/10.1038/s41586-026-10656-8