AlphaFold protein structure prediction

A protein's job depends on its shape. Proteins are chains built mostly from 20 kinds of amino acids that fold into three-dimensional structures, and since the 1970s researchers had tried to predict those structures from the amino acid sequence alone, a notoriously difficult task. The Royal Swedish Academy of Sciences later called it a 50-year-old problem.

In 2020 Demis Hassabis and John Jumper of DeepMind in London presented AlphaFold2, a deep-learning system that built physical and biological knowledge about protein structure into its design. In the 14th Critical Assessment of protein Structure Prediction, CASP14, it produced structures with accuracy competitive with experimental structures in a majority of cases and greatly outperformed other methods. Jumper, Richard Evans and 32 co-authors, with Hassabis as senior author, described the method in Nature, published online on 15 July 2021, as the first computational method that could regularly predict protein structures with atomic accuracy even when no similar structure was known. Earlier methods, they wrote, fell far short of atomic accuracy, especially when no related structure was available.

The model was then used to predict the structure of virtually all of the 200 million proteins researchers have identified. By 2024 more than two million people in 190 countries had used AlphaFold2, for work that included studying antibiotic resistance and imaging enzymes able to break down plastic.

On 9 October 2024 the Royal Swedish Academy of Sciences awarded half of the Nobel Prize in Chemistry jointly to Hassabis and Jumper for protein structure prediction. The other half went to David Baker of the University of Washington for computational protein design.