Machine Learning Enables Universal Protein Activation in Living Mice

In a major advancement for biomedical research and potential therapies, scientists have developed a powerful new tool to control protein activity inside living animals with unprecedented precision. The method, called CAGE-Proxvivo, allows researchers to activate or deactivate specific proteins in live mice using a small molecule trigger—offering a flexible and targeted approach to study how proteins function in real time.

Developed with the help of machine learning, the technique involves engineering special amino acids that are chemically "caged," temporarily blocking a protein's function. These caged amino acids are inserted at specific sites within the protein, and their blocking effect can later be reversed inside the body using a bioorthogonal chemical reaction—essentially flipping a molecular switch to turn the protein back on.

This system enables not only protein activation but also precise control over protein-protein interactions. For example, researchers demonstrated a "gated" antibody that activates T cells only when triggered by a chemical signal—potentially enhancing immune responses specifically at tumor sites.

Published as a proof of concept in live mice, CAGE-Proxvivo represents a versatile platform for studying protein function in complex biological systems and could one day lead to more controlled and personalized therapeutic strategies. By combining machine learning with chemical biology, the research opens the door to time-resolved experiments and tightly regulated treatments within living organisms.

SOURCE: https://www.cell.com/cell/abstract/S0092-8674(25)00517-3 

CREDITS: CELL PRESS