AlphaFold Five Years On: Protein Structure Predictions Used by Over 3 Million Researchers

In a November 2025 post marking five years of AlphaFold, Google DeepMind's AlphaFold team described how the system moved from a research result to a standard scientific tool. At the CASP14 competition in 2020, AlphaFold 2 predicted protein structures from their amino acid sequences with an accuracy widely hailed as a solution to a 50-year-old challenge in biology; determining a single structure experimentally can take a year or more of expensive work. In 2021 DeepMind launched the AlphaFold Protein Database in partnership with EMBL-EBI, and a year later it released AlphaFold 2 predictions for more than 200 million protein structures. The freely available database has been used by over 3 million researchers in more than 190 countries, including over 1 million users in low- and middle-income countries, and over 30% of AlphaFold-related research focuses on understanding disease. The work was recognised with the 2024 Nobel Prize in Chemistry. DeepMind cites examples such as European scientists studying Vitellogenin, a key immunity protein in honeybees, to support conservation and breeding programmes; AlphaFold 2 helping reveal the structure of apolipoprotein B100, the central protein in LDL cholesterol; and plant researchers at the University of Zurich and the Sainsbury Lab using it alongside comparative genomics. AlphaFold has been cited in more than 35,000 papers, and more than 200,000 papers incorporated elements of AlphaFold 2 in their methods. An independent analysis by the Innovation Growth Lab found that researchers using AlphaFold 2 submit over 40% more novel experimental protein structures, and that linked research is twice as likely to be cited in clinical articles. Isomorphic Labs, an AI drug discovery company founded in 2021, developed AlphaFold 3 with DeepMind. AlphaFold 3 predicts the structures and interactions of proteins, DNA, RNA and ligands, and the AlphaFold Server, which offers it to non-commercial researchers, has made more than 8 million predictions. DeepMind has since built related models such as AlphaMissense, AlphaGenome and AlphaProteo.