The Nobel Prize in Chemistry for protein structure prediction
Half the prize went to David Baker for protein design, and half to Demis Hassabis and John Jumper for AlphaFold.
Why it matters
The award went not to a discovery in nature but to a program, and for the second day running machine learning took a Nobel.
The protein folding problem had stood for fifty years: the sequence of amino acids was known, and the shape did not follow from it. AlphaFold gave predictions of accuracy comparable to experiment, and a database of two hundred million structures was opened to everyone. Baker solved the inverse problem, designing a sequence that folds into a given shape, and did it before deep learning, later carrying it over. Two prizes in two days meant that methods which had lived in computer science for decades had become an instrument of the natural sciences.