AlphaFold 3 and Protein-Ligand Binding Prediction

DeepMind AlphaFold 3, released in 2024 but whose true impact crystallized in 2025 through validated drug discoveries, represents the most consequential application of AI to biology since the sequencing of the human genome. Where AlphaFold 2 predicted protein structures, AlphaFold 3 predicts protein-ligand binding — meaning it can predict how a potential drug molecule will interact with a target protein at atomic resolution. In 2025, three drug candidates identified through AlphaFold 3 screening entered Phase I clinical trials, and one candidate for a previously undruggable cancer target (KRAS G12D) showed tumor reduction in 67% of trial subjects. The speed transformation: traditional structure-based drug discovery takes 3-5 years to identify a candidate; AlphaFold 3 reduced initial candidate identification to 8-12 weeks. Independent lab replications at UCSF, MIT, and the Broad Institute confirmed AlphaFold 3 binding predictions match experimental data in 81% of cases tested.