AlphaFold 3
DeepMind's model predicting the structure of proteins, DNA, RNA, and ligands together
AlphaFold 3 is a structure-prediction model that Google DeepMind and Isomorphic Labs released in May 2024, extending the earlier protein-only AlphaFold 2 to a much wider range of biological molecules. Instead of predicting only protein folds, it jointly models complexes containing proteins, DNA, RNA, small-molecule ligands, ions, and chemically modified residues, using a redesigned architecture built around a module called Pairformer. This lets researchers ask how a candidate drug molecule might bind to a protein, or how a protein interacts with a strand of RNA, in a single prediction rather than stitching together separate tools.
DeepMind made the model available through the free AlphaFold Server for non-commercial research, and in November 2024 released the underlying code and weights for academic use as well, though commercial use still runs through Isomorphic Labs. AlphaFold 3 is widely used in structural biology and early-stage drug discovery, and it extended DeepMind’s earlier Nobel Prize-winning work on protein structure prediction into the broader space of molecular interactions that drive most drug design and disease research.