AlphaFold Opens a Free Global Library of Protein Shapes

AI-powered AlphaFold2 transformed protein folding by predicting 3D structures from sequence in minutes, leading to a public database with roughly 214 million predicted protein structures available for free. Most entries are highly accurate or confidently useful, but about a fifth have low confidence and users should filter by pLDDT. The data is CC-BY-4.0 licensed and widely used, enabling rapid research while raising questions about openness for future commercial tools and the handling of multiple conformations and disordered regions. Ongoing updates include isoforms and alignments; methods like AF-Cluster have had mixed results in extracting alternative conformations.