Nature 510, 103–108 (2014)

Inspired by biological systems, researchers have sought to use proteins to create synthetic self-assembled nanostructures. Computational methods have, in particular, been developed that allow novel protein assemblies to be built with atomic-level precision, but the structures designed so far have been relatively basic and use only one type of building block. David Baker and colleagues at the University of Washington and the University of California, Los Angeles have now developed a computational method that allows nanostructures to be designed from two distinct subunits.

The approach is based on Rosetta software, which was originally developed by Baker and colleagues to predict the structure of naturally occurring proteins. First, the docking of protein building blocks on defined symmetry axes is systematically explored and used to identify large interfaces with high densities of contacting residues. Then, sequences of amino acids are designed to stabilize these interactions and drive co-assembly of the two components.

The team used the method to design five different cage-like nanostructures, which each contained 24 subunits. The structures were synthesized and, with the help of electron microscopy and X-ray crystallography, shown to closely resemble the computational models.