Researchers at Google DeepMind have published a landmark paper in Nature describing AlphaFold 3, a new version of their protein structure prediction system that can accurately model how proteins interact with each other and with small molecules. The advance represents a significant step toward computationally driven drug discovery.
AlphaFold 3 can predict protein-protein interaction complexes with atomic-level accuracy, a capability that was previously limited to expensive and time-consuming experimental methods like X-ray crystallography and cryo-electron microscopy. The system can model an interaction complex in minutes that would traditionally take months of laboratory work.
Pharmaceutical companies are already licensing the technology to accelerate their drug development pipelines. Novartis announced it will use AlphaFold 3 to identify potential drug targets for neurodegenerative diseases, while Pfizer plans to apply the system to antibody engineering for cancer immunotherapies.
