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.