This is a sample archive article demonstrating how older content appears in our back catalogue.
On May 8, 2024, Google DeepMind published AlphaFold 3 in the journal Nature. The system predicts the structure and interactions of all life's molecules with unprecedented accuracy.
What AlphaFold 3 Actually Does
Previous versions predicted protein structures. AlphaFold 3 expands to DNA, RNA, ligands, and post-translational modifications. It models how these molecules interact inside cells.
Verified Performance Data
- Protein-Ligand Complexes: 76% accuracy on PoseBusters benchmark (next best tool: 52%)
- Protein-Nucleic Acid: 65% accuracy on recent PDB structures
- Antibody-Antigen: 62% accurate in docking predictions
Source: Abramson et al., Nature 630, 493-500 (2024). Peer-reviewed.
Real-World Impact
The AlphaFold Database contains over 200 million protein structures. Scientists have used it for:
- Malaria vaccine development (University of Oxford, 2024)
- Plastic-degrading enzyme engineering (University of Texas, 2023)
- Crop yield improvement (Corteva Agriscience, 2024)
Limitations (As Stated by DeepMind)
- Confident predictions require multiple sequence alignments
- Dynamic protein movements are not fully captured
- Some complexes remain outside training distribution
AlphaFold Server is free for non-commercial research. Isomorphic Labs (DeepMind spinout) holds commercial rights.
Sources
- Abramson, J. et al. "Accurate structure prediction of biomolecular interactions with AlphaFold 3." Nature 630, 493-500 (2024). DOI: 10.1038/s41586-024-07487-w
- DeepMind Blog, "AlphaFold 3 predicts the structure and interactions of all life's molecules," May 8, 2024
- European Bioinformatics Institute, AlphaFold Database statistics, accessed May 2026
This article was originally published May 28, 2026. All data verified against primary sources.