Protein structure prediction by deep learning (AlphaFold 2)
Predicción de la estructura de proteínas por aprendizaje profundo (AlphaFold 2)
- Date
- 2020-2021 · confidence high
- Place
- DeepMind, London, England
- Credited to
- Demis Hassabis and John Jumper (DeepMind); David Baker (protein design, Seattle)
- People
- 👤 David Baker1962–👤 Demis Hassabis1976–👤 John Michael Jumper1985–
- Field
- Life & health
What was discovered
AlphaFold 2 predicts the three-dimensional structure of almost any protein from its sequence with experimental accuracy, solving a 50-year-old problem; DeepMind publishes 200 million structures. Baker demonstrates the design of new proteins.
Nobel Prize in Chemistry 2024 (Baker, Hassabis and Jumper).
What it led to
- Structure-based drug design at scale
- Computational structural biology
- Design of de novo enzymes and proteins
- AI in chemistry (AlphaFold 3 for complexes, 2024)
Built on
- Backpropagation for training neural networks1986
- Transformer architecture2017
- protein crystallography (Kendrew, Perutz)
- Levinthal's paradox and CASP
Made possible
- databases of predicted structures
- designed enzymes