Nordic Life Science 1
48 “How the proteins fold into very specific shap
es has been a mystery, but the laureates have opened doors to this universe and given us tools that help us understand how proteins are built and turned into machines with distinct tasks.” of the Institute for Protein Design at University of Washington School of Medicine that is often referred to as a start-up factory,” he adds. From Rosetta to AI In 2003 Professor David Baker and his colleagues published the design and crystallographic validation of a 93residue α/β-protein named Top7 – a breakthrough in computational de novo protein design. De novo protein design means the generation of new proteins with sequences unrelated to those in nature based on physical principles of intramolecular and intermolecular interactions. This was big news for several reasons; it was a relatively large protein, they sought to design a folding pattern not found for any globular protein in the Protein Data Bank and the final design had no significant sequence similarity to any naturally occurring protein in the sequence databases. As such, Top7 was unlike any other protein, both structurally and sequence-wise, and it was designed by automated computation with full optimization of both backbone and sidechains. The Royal Swedish Academy of Sciences’ describes that the key to Baker’s success was his team's initial development of the Rosetta computer program in 1999. The program assembles short structural fragments from unrelated protein structures with similar local sequences in the Protein Data Bank and simultaneously optimizes sequence and structure with respect to the target backbone conformation. Baker and his colleagues showed that a wide range of protein structures could be designed using the Rosetta software, including proteins that can be used as pharmaceuticals and vaccines. In addition, in 2008, Baker and his coworkers reported the first attempts at de novo enzyme design, and the design of novel enzymes that can catalyze reactions for which no naturally occurring enzymes exist. T he second discovery awarded with a Nobel Prize this year concerns the prediction of protein structures. Ever since the 1970s researchers have had difficulties predicting protein structures from amino acid sequences. The breakthrough came just four years ago, in 2020, when Demis Hassabis and John Jumper, DeepMind, presented an artificial intelligence (AI) model, AlphaFold2. AlphaFold2 (AF2) showed a level of accuracy that was competitive with experimental structures for a majority of target proteins. “The AF2 architecture can be described as an ingenious piece of neural network with a multitude of new inventions, and it can be viewed as the first real scientific breakthrough of artificial intelligence,” stated the Royal Swedish Academy of Sciences, emphasizing the impact of this development. The AF2 source code was also made public which decisively contributed to its impact as it could be extensively tested and validated. Since 2020, more than two million people from 190 countries have used it. Crucial tools Thanks to the laureates’ protein design and prediction tools, today we can easily visualize the structure of proteins THE NOBEL PRIZE // CHEMISTRY