Researchers use AI to design proteins that cut up other proteins

Proteins are built from amino acids linked by peptide bonds, and those bonds are stubbornly stable on their own. In the environment, though, proteins usually don’t stick around for centuries. Living things have spent a long time evolving enzymes that can break them apart, and that family has splintered into all kinds of specialized proteases.

The new work skips that slow evolutionary sprawl. Researchers used AI-based protein design tools to keep the catalytic core from two protease types and replace the rest of each protein. That is a pretty dramatic edit. Instead of taking a natural enzyme and nudging it, they essentially rebuilt most of it from scratch.

The result was a set of enzymes that could be tuned to cut chosen proteins in half. To show they worked, the team used them against proteins tied to neurological diseases. That’s the useful bit here: not just that the enzymes cut, but that they could be aimed.

It is an elegant shortcut. Nature’s proteases got to their variety by trial, error, and a lot of time; this approach starts from a known catalytic engine and lets AI design the surrounding protein around the job you want done. The source doesn’t claim this has solved protein degradation wholesale. It does show that AI can help compress a process that evolution usually handles in pieces.