Artificial intelligence has solved a 50-year-old biological mystery by predicting the 3D structures of nearly all known proteins.

For over five decades, scientists grappled with the protein folding problem: determining a protein's 3D shape from only its amino acid sequence. This challenge was first formally identified in 1972 by Nobel laureate Christian Anfinsen. Traditional experimental methods like X-ray crystallography and cryo-electron microscopy are expensive, time-consuming, and can require years of laboratory work to determine a single protein's structure.In 2020, Google DeepMind achieved a major breakthrough during the CASP14 competition with AlphaFold 2, demonstrating the ability to predict protein structures with accuracy matching experimental methods. By 2022, DeepMind and the European Bioinformatics Institute (EMBL-EBI) released a comprehensive database containing predictions for nearly all 214 million known proteins, covering humans, plants, bacteria, and even extinct species.The AI uses a deep learning architecture inspired by how the human brain processes information to model the physical forces between atoms. This technology has already enabled researchers to design new enzymes capable of breaking down plastic waste and to identify potential vaccine targets for malaria. In essence, it compressed centuries of manual research into just a few years of computation.