AI-designed proteins work as molecular sensors in living cells
Researchers at Queensland University of Technology have created artificial proteins that act as switches, activating only when they detect a specific target molecule. The work, published in Nature Biotechnology, demonstrates that machine learning can design proteins suitable for biosensors in medicine, environmental monitoring and biotechnology.
The team built protein receptors using AI and connected them to enzymes that produce measurable outputs: colour changes, light emission, or electrical signals. They tested the switches against small molecules, peptides and proteins, and showed they function inside living bacterial cells and can link to electrodes like a glucose meter.
How the switches work
When a target molecule binds to the AI-designed receptor, it subtly alters how the protein moves. That physical change is enough to activate the connected enzyme and trigger a detectable signal.
This finding challenges a long-held assumption in protein science. Researchers previously believed sensing proteins required dramatic shape changes to function as switches. The new work shows that small shifts in protein dynamics are sufficient.
Practical applications emerging
The researchers demonstrated electrochemical biosensors for steroid detection and confirmed the switches work in living cells. These results suggest the technology could support portable diagnostic devices, environmental monitoring systems and engineered cells that respond to chemical signals.
The international collaboration included seven research teams from Australia, the UK and the US. The work involved researchers from the University of Washington, led by 2024 Nobel Prize laureate David Baker, and CSIRO.
For professionals working in AI for Science & Research, this demonstrates how machine learning systems can generate novel molecular solutions that expand what protein engineers can build.
Your membership also unlocks:
