Can We Design a Protein That Has Never Existed?
We already have. The harder question is whether we can make new ones that do useful work.
By The Editors/October 24, 2026/1 min read
In 2003 a team in David Baker's lab reported Top7, a 93-residue protein with a fold not seen in nature. The designed structure and the solved crystal structure matched closely. It did nothing in particular. Its achievement was to exist.
Twenty years later, the tools look different. RFdiffusion, published in 2023, generates protein backbones with a method related to the diffusion models behind AI image generators. Designers can ask for a binder against a chosen target or a symmetric assembly, then test candidates in the lab at a scale that used to be impractical.
Function is the hard part
Designed proteins have already left the lab. SKYCovione, a COVID-19 vaccine built on a computationally designed nanoparticle, was approved in South Korea in 2022.
Enzymes remain a much harder target. Natural catalysts depend on dynamics and precise chemistry that a static model captures poorly, and evolution has had a very long time to tune them. A design that folds correctly can still be a mediocre catalyst.
The most interesting labs now are the ones treating design and experiment as a single loop, where failures feed the next round of models. The space of proteins nature never tried is unimaginably large. We are only starting to learn which parts of it are worth visiting.
Sources
- Kuhlman B et al. Design of a novel globular protein fold with atomic-level accuracy. Science, 2003.
- Watson JL et al. De novo design of protein structure and function with RFdiffusion. Nature, 2023.
- South Korea Ministry of Food and Drug Safety approval of SKYCovione (GBP510), SK bioscience with the Institute for Protein Design, June 2022.
