Protein structure prediction (protein folding problem)

Settled in 2020 by DeepMind's AlphaFold2 team (Demis Hassabis, John Jumper, and colleagues).

Christian Anfinsen's experiments in 1961 showed that a protein's amino acid sequence alone determines its final folded 3D shape (earning him the 1972 Nobel Prize), but predicting that shape computationally from sequence alone proved to be a fifty-year grand challenge in biology, tracked since 1994 by the biennial CASP competition. In the 2020 CASP14 competition, DeepMind's AlphaFold2 achieved median accuracy competitive with experimental methods for most target proteins, a result the CASP organizers themselves described as effectively solving the core sequence-to-structure prediction problem; DeepMind subsequently released predicted structures for nearly all known proteins via the AlphaFold Database. Demis Hassabis and John Jumper shared half of the 2024 Nobel Prize in Chemistry for this work (with David Baker sharing the other half for computational protein design).

A problem biologists spent half a century calling nearly impossible was cracked by an AI system trained to play a very different kind of game, turning the prediction of life's molecular machinery from a laboratory bottleneck into an instant calculation.

Explanations that circulated

• That AlphaFold's success means every aspect of protein folding is now understood — it predicts static folded structures with remarkable accuracy but does not fully solve the separate, related questions of folding dynamics/kinetics (how a protein folds moment-to-moment, the 'Levinthal paradox' route) or structure prediction for highly flexible/disordered proteins and complex multi-protein assemblies, which remain active research areas

References

• Highly accurate protein structure prediction with AlphaFold, John Jumper, Richard Evans, Alexander Pritzel, et al., Nature, 2021
• The Nobel Prize in Chemistry 2024, The Royal Swedish Academy of Sciences, Nobel Prize Press Release, 2024