AI untangles ‘the protein-folding problem’
One of the biggest mysteries in biology may have been solved by an artificial intelligence program. The London-based DeepMind lab, owned by the same parent company as Google, says its new computer system can successfully predict how a protein folds into a unique 3D shape. The “protein-folding problem” has befuddled scientists for half a century.
Its importance lies in the fact that most biological processes—such as how insulin controls blood sugar levels or how antibodies fight coronaviruses—are driven by proteins. How the strings of amino acids that make up a protein are twisted and folded determines its function. Trying to establish exactly how proteins get their origami-like structure is fiendishly complex and can take years of lab work. That’s why scientists have solved the shape of only about 170,000 of the 200 million proteins identified so far. But at a recent protein-folding competition for computer scientists, DeepMind’s new AI, AlphaFold, showed it can solve a protein’s puzzle in a matter of hours with an accuracy level similar to that of lab work. Experts caution that this technology will speed up only one part of the long process of developing medicines and analyzing diseases. But the AI is already being used for work on malaria, sleeping sickness, and a parasitic disease called leishmaniasis. Unraveling protein structures could “really help us understand how human beings operate and function,” Janet Thornton, from the European Bioinformatics Institute, tells The Guardian (U.K.). “This is a problem that I was beginning to think would not get solved in my lifetime.”