Proteases are enzymes that cut proteins apart, and nature has evolved many versions with different levels of specificity. Now researchers have used AI-based protein design tools to shortcut that evolutionary process. Starting from the catalytic cores of two existing proteases—a zinc protease from crayfish and a cysteine protease from yeast—they generated entirely new surrounding backbones with RFdiffusion2, then populated those backbones with amino acids using ProteinMPNN.
The design constraints forced each new backbone to bring the key catalytic amino acids together while also accommodating a target protein sequence near the active site. This was a demanding task because peptide bonds are extremely stable without a catalyst. The team ran the process dozens of times, testing 12 different backbones and 69 amino-acid sets for the cysteine protease, and 135 separate designs for the zinc protease. AlphaFold3 was then used to predict the structures of many of those designs.
According to Ars Technica, the resulting enzymes could be tuned to cut specific proteins in half, and the researchers demonstrated this by using them to cut proteins associated with neurological diseases. The work is an early demonstration of custom-built proteases, though the source does not describe any clinical applications yet.