De novo protein design

"De novo" is Latin for "anew" or "from scratch". In de novo protein design, researchers therefore do not make a copy or variant of a protein from nature, but invent an entirely new protein for a task they choose.

How does it work?

A protein is a chain of building blocks, called amino acids, that folds into a fixed shape. That shape determines what the protein can do, for example attach itself to a particular target or convert one substance into another. In de novo design the reasoning runs the other way round: first it is decided which shape and which properties the protein should have, and then a sequence of building blocks is sought that folds into that shape.

That is difficult, because there are an unimaginable number of possible sequences and only a small share of them lead to a usable shape. Computer models help to find good candidates for a desired shape. Nowadays these are mainly AI models that have learned from the proteins already known. The candidates are then tested in a lab to see whether they work, because a design on the computer is not yet a working protein.

Why it matters

Nature has an enormous collection of proteins, but there is not a protein for every problem. With de novo design, researchers can make proteins that nature has not come up with, for example for medicines or for neutralising substances.

That AI is making strong progress here could be seen in WATTBA's news. Chinese researchers built an AI model that designs proteins from a description in plain language, and in four trials the result really did what was asked: give off light, break down plastic, convert alcohol and help capture CO2 (link). The lab of Nobel laureate David Baker took AI in a different direction and designed enzymes that cut other proteins at a pre-chosen spot, including the protein that clumps together in Alzheimer's disease (link).

These are research results. It is still a long road of testing and evaluation before they lead to a new medicine or application.