Two attempts. The first failed, the second succeeded. >When Baker and his team applied this second network to their hallucinated protein nanoparticles, it had

Two attempts. The first failed, the second succeeded. >When Baker and his team applied this second network to their hallucinated protein nanoparticles, it had much greater success making the molecules experimentally. The researchers determined the structure of 30 of their new proteins using cryo-electron microscopy and other experimental techniques, and 27 of them matched the AI-led designs2. The team’s creations included giant rings with complex symmetries, unlike anything found in nature. In theory, the approach could be used to design nanoparticles corresponding to almost any symmetric shape, says Lukas Milles, a biophysicist who co-led the effort. “It is electrifying to see what these networks can do.