AlphaFold is an artificial-intelligence system built by DeepMind that predicts the three-dimensional shape of a protein from the sequence of amino acids that make it up. That shape matters because a protein's job is determined almost entirely by how it folds: get the structure, and you can begin to understand the function, and from there the disease. For roughly fifty years, working out these shapes was painstaking laboratory work that could take months or years for a single protein.
In 2020, AlphaFold2 reached an accuracy that rivalled those slow physical methods, and within a few years the system had predicted the structures of around two hundred million proteins — very nearly every one known to science. DeepMind released the database free of charge. In 2024, the work earned Demis Hassabis and John Jumper a share of the Nobel Prize in Chemistry, alongside David Baker for the related field of protein design.
What makes AlphaFold significant beyond biology is what it demonstrated about AI itself. A problem that the field had quietly accepted might never fully yield was not gradually worn down but largely dissolved, by a machine that learned the underlying grammar of folding from existing examples. It stands as one of the clearest cases of an AI system delivering a genuine scientific breakthrough rather than an incremental improvement — a fact in the past tense, not a promise about the future.
One honest limit is worth keeping in view. Predicting a structure is not the same as understanding everything a protein does, how it moves, or how to turn that knowledge into a working medicine. AlphaFold compressed one slow step in a long chain. It did not collapse the whole chain — but compressing even one step of that magnitude reshaped what the rest of the field could attempt.
