2021: AlphaFold's structures are published in Nature

On 15 July 2021, DeepMind and collaborators published a landmark paper in the journal Nature titled "Highly accurate protein structure prediction with AlphaFold." It described how a deep learning system could predict the three-dimensional shape of a protein from its amino acid sequence with a level of accuracy that rivaled years of painstaking laboratory work using X-ray crystallography or cryo-electron microscopy.

The problem AlphaFold tackled, often called the "protein folding problem," had frustrated biologists for half a century. Proteins are built from chains of amino acids, but it is their folded 3D structure that determines how they function in the body. Knowing that structure is essential for understanding diseases and designing new drugs, yet determining it experimentally could take months or even years per protein.

AlphaFold's abilities had first been demonstrated publicly at the CASP14 competition in December 2020, where it dramatically outperformed rival methods. The July 2021 Nature paper, published alongside a companion paper in the same issue, gave the scientific community the full technical details and, crucially, was accompanied by the release of an open database of predicted structures, developed with the European Bioinformatics Institute. Researchers around the world could now look up predicted structures for proteins they were studying instead of waiting years for experimental data.

• The database has since grown to cover over 200 million proteins, essentially the majority of known proteins in science.
• The work has been used in research on antibiotic resistance, malaria vaccines, plastic-eating enzymes, and much more.
• In 2024, Demis Hassabis and John Jumper shared the Nobel Prize in Chemistry for this achievement, alongside David Baker for related protein design work.

AlphaFold is often cited as one of the clearest examples of AI producing genuine scientific value beyond text and images. It shows how machine learning, trained on large curated datasets, can compress decades of specialized experimental labor into seconds of computation. As AI tools increasingly move into chemistry, materials science, and medicine, 15 July 2021 stands as a useful reminder that some of the technology's most lasting contributions may be quietly reshaping laboratories rather than headlines.