A newly developed computational approach makes it possible to design peptides with defined three-dimensional structures that could be useful for drug discovery (Nature 2016, DOI: 10.1038/nature19791). Peptides are often too floppy and undisciplined to interact with molecular targets in a consistent way. But the designed peptides’ disulfide bonds and terminal cyclization constrain them into defined shapes that enhance their potential for bioactivity. Protein design generally relies on prior knowledge of natural protein structures, but first-principles simulation techniques reported in the new study by David Baker of the University of Washington and coworkers optimize the creation of nonnatural peptides without reference to known structures. They used the system to design 16 peptides with unprecedented shapes, in some cases made from both natural and nonnatural amino acids. They created the designed structures by peptide synthesis or by expression in engineered bacteria. The designed peptides are resistant to thermal and chemical denaturation, and X-ray and NMR structures of 12 of them showed them to be nearly identical to their computational design models. “We are now working on developing peptides that target pharmacologically relevant proteins,” says Baker group postdoc Gaurav Bhardwaj.