AI Helps Scientists Find an Enzyme That Repairs a Hallmark of Aging

Artificial intelligence is changing the inventing process in ways that are easy to miss.

Much attention has focused on AI writing code, generating images, or answering questions. But one of AI's most important roles may be helping invent entirely new things that humans would be unlikely to discover on their own.

A 2026 study in Nature Communications provides an excellent example. Researchers engineered an enzyme called CMLase to remove CML, a widespread form of age-related chemical damage found on long-lived proteins throughout the body, and showed it could reduce that damage in human tissue samples. Rather than trying to slow the aging process, the enzyme actually removes accumulated chemical damage and restores proteins closer to their original state.

What's remarkable isn't just the enzyme. It's how they invented it.

Searching Where Humans Can't

Proteins are incredibly complex. A tiny change in their structure can make the difference between success and failure. The number of possible designs is so enormous that no scientist could ever evaluate them one by one.

The team used AI structural predictions to narrow the field before any lab work began. Pulling from a public database of AI-predicted protein structures, they screened tens of thousands of candidate enzymes and identified fewer than fifty with a specific shape likely to grab onto their target. AI didn't design the enzyme from scratch, it helped the researchers find the right starting point in a haystack too large to search by hand.

From there, the real workhorse was a much older technique: directed evolution. The researchers engineered bacteria so they could only survive and grow if the enzyme actually worked, then let that survival pressure do the sifting - generation after generation, mutation after mutation. Over five rounds of this process, they evaluated more than 500 million enzyme variants, arriving at one capable of removing CML, one of the most common forms of age-related protein damage.

The result wasn't generated by AI alone. It was created through a partnership between human scientists, AI-assisted structural search, and old-fashioned biological trial and error running at a scale no human could match.

A New Model for Inventing

For inventors, this is the bigger story.

Traditionally, invention has meant starting with an idea and gradually refining it through experience, intuition, and trial and error. AI introduces another approach. Instead of searching through a handful of ideas by hand, inventors can now use AI to point them toward the most promising starting places among millions of possibilities.

That doesn't replace creativity. It amplifies it.

The researchers still had to identify the problem, understand the biology, recognize what success looked like, and prove the result experimentally. AI simply helped them aim their search in a design space that was far beyond human reach.

The Lesson

The best inventors are the ones who ask different questions. This team asked a powerful one:

"What if damaged proteins could be repaired instead of simply protected?"

Then they combined AI and biology to go find an answer.

That's an important shift. AI isn't just becoming a better assistant. It's becoming one tool among several in the modern inventor's kit - one that allows inventors to explore ideas that would have been practically impossible just a few years ago.

As these tools improve, we'll likely see breakthroughs not because AI is replacing inventors, but because inventors are learning how to combine AI with the right complementary tools to discover solutions neither could have found alone.

Invention City works with inventors across a range of biotech and medtech opportunities. We're currently evaluating an exciting invention that addresses issues emerging around GLP-1 use. If you have a research-backed invention related to gastric motility, we'd like to hear from you. Contact Mike Marks directly at mike@inventioncity.com.