Protein structure prediction reached a new level

Abstract

Structural and computational biologists made tremendous advances this past year using artificial intelligence techniques, like neural networks, to predict the way proteins fold. Methods that can take a protein sequence and predict how it will fold are critical to helping scientists better understand biological systems, including guiding drug developers to design molecules to hit specific targets. But more importantly, experts say this boom in protein structure prediction is the beginning of something, not the end. The challenge of predicting protein shape and geometry is in taking the primary structure of proteins—the long list of individual amino acids—and spitting out a reliable 3D structure. There are 20 natural amino acids , each with a different side chain and therefore unique chemistry. These units interact with one another and the surrounding environment, causing protein strands to shimmy into various functional shapes, including small peptides that act as signaling molecules and large, complicated