How did AlphaFold solve a 50-year-old biology grand challenge?

Predicting how proteins fold was biology's holy grail. AI cracked it in a weekend.

The answer

In December 2020, DeepMind's AlphaFold 2 achieved a median GDT score of 92.4 out of 100 at the CASP14 competition, effectively solving the protein structure prediction problem that had stumped biologists for over 50 years. In July 2022, DeepMind released predicted structures for nearly all 200 million known proteins, a dataset that would have taken experimental methods an estimated billion years to produce. The system uses a novel architecture combining attention mechanisms with evolutionary and physical constraints. The breakthrough earned DeepMind co-founder Demis Hassabis and John Jumper the 2024 Nobel Prize in Chemistry.

Source: Nature

More from AI & Machine Learning

All AI & Machine Learning →