Drug design has long been compared to fitting pieces into a complex jigsaw: each protein target has a unique shape and pockets where drugs can latch on. Traditional drug development has excelled at targeting well-folded, stable proteins—but many critical proteins, especially those implicated in cancer and Alzheimer’s, are intrinsically disordered, shifting constantly and defying conventional targeting methods.
A team from the University of Washington led by David Baker has now changed the game. They’ve developed a novel AI tool that can recognize these elusive, wobbly proteins and generate “binders”—custom-designed molecules that attach even to shifting targets.
Their process combines pattern recognition of recurring structures, assembly of diverse binder templates, and refinement through advanced AI techniques. The result? About a thousand potential binding solutions capable of handling trillions of combinations.
As proof of concept, the team designed binders for 39 different disordered proteins. Notably, they targeted dynorphin A, a dynamic protein involved in pain signaling. Remarkably, these AI-crafted binders locked onto dynorphin A more effectively than its natural binding partners and blocked pain signaling in lab-grown human cells.
This approach marks the potential birth of a whole new class of medicine—one that can clamp onto previously untouchable proteins. It opens doors to influencing cellular structures like biomolecular condensates, which regulate gene expression and immune responses, and it paves the way for binder-based therapies that mimic antibodies. Beyond treatment, these tools could revolutionize synthetic biology by enabling exploration of dynamic protein interactions in living cells.
While extensive testing remains—particularly around safety and durability—the breakthrough offers a promising glimpse into a future where even the most elusive disease targets can finally be drugged.