Researchers have long struggled to design small molecules that can target proteins with unstable structures, but two new artificial intelligence-based systems could change the game. A preprint published by Kejia Wu and her colleagues reveals a deep learning-based approach that uses flexible binders to target disordered proteins.
Intrinsically disordered regions are common in many proteins, making them difficult to target. Unlike stable proteins, which have fixed structures, these proteins can change shape rapidly, making it hard to find a molecule that binds effectively.
The new method, developed by Wu and her team, uses an amino acid sequence as input to design a flexible binder. This binder can adapt to the changing structure of the protein, allowing it to bind more effectively. The approach has been shown to work well for cases where the disordered protein can adopt some helix or strand conformation.
Building on this idea, a new technique called “logos” has been developed. Logos creates its own pocket that the target protein fits into, rather than trying to fit into an existing pocket. This approach works best for proteins that don’t want to be in a specific structure, and could lead to breakthroughs in treating diseases such as cancer and neurodegenerative disorders.
One area where these binders could have a significant impact is in the treatment of Alzheimer’s disease. The intrinsically disordered protein tau has been implicated in the disease, and flexible binders like those developed by Wu’s team could disrupt its damaging aggregation. With the increasing prevalence of neurodegenerative diseases in modern society, this technology could bring new hope for patients suffering from these conditions.
Source: https://cen.acs.org/articles/103/web/2025/07/New-AI-assisted-methods-take.html