AlphaFold

AlphaFold is DeepMind's AI system that solved protein structure prediction—determining the 3D shape of a protein from its amino acid sequence—a problem that had stumped biology for 50 years. Its success earned Demis Hassabis and John Jumper the 2024 Nobel Prize in Chemistry and stands as perhaps the most consequential application of AI to science.

Proteins are the molecular machines of life, and their function is determined by their 3D shape. But predicting how a chain of amino acids folds into a specific structure is extraordinarily complex—the number of possible configurations for even a small protein is astronomically large. Experimental methods (X-ray crystallography, cryo-electron microscopy) take months to years per protein and cost tens of thousands of dollars each. AlphaFold predicts structures in minutes with accuracy rivaling experimental methods.

AlphaFold 2, released in 2020, predicted the structure of nearly every known protein—over 200 million structures—and made the database freely available. AlphaFold 3 (2024) extended to predicting interactions between proteins, DNA, RNA, and small molecules, directly relevant to drug design. The impact on biomedical research has been transformative: thousands of studies now use AlphaFold predictions to understand disease mechanisms, design drugs, and engineer new proteins.

AlphaFold is the flagship example of AI for scientific discovery—demonstrating that deep learning can solve problems that domain experts couldn't, not by following human approaches faster but by learning fundamentally different representations of the underlying physics. The same principle applies in materials science, climate modeling, and mathematics. Like AlphaZero for games, AlphaFold proved that AI can discover knowledge inaccessible to human cognition alone.