protein prediction database

John Jumper describes how AlphaFold uses machine learning to predict protein structures, helping researchers investigate biology and develop new scientific hypotheses. He explains that its progress came from combining public protein structure data, compute, and many research ideas, with the research contributing far more than simply scaling the data. Blind assessments helped show that AlphaFold was a major advance, while open code and a public prediction database let biologists test it on their own work and build trust. He closes with examples of scientists using AlphaFold in unexpected ways, arguing that AI tools amplify experimental science and may lead to broader scientific capabilities.