Textbook of AI

AlphaFold 2's CASP14 victory in December 2020 is generally regarded as the moment at which the protein-folding problem, predicting 3D structure from amino-acid sequence, was effectively solved. CASP (Critical Assessment of Structure Prediction) is the biennial blind protein-structure-prediction competition that the structural-biology community has run since 1994; CASP14 ran from May to August 2020, with results announced in November-December.

AlphaFold 2 (developed at DeepMind by John Jumper, Demis Hassabis and colleagues) achieved median GDT-TS scores around 92.4 (out of 100) on hard targets, a dramatic improvement over previous methods including AlphaFold 1 (CASP13 2018) and over experimental accuracy thresholds. CASP organisers Andrei Lupas and others stated publicly that "in some sense, the problem is solved", while noting remaining edge cases.

The result transformed structural biology overnight. The 2021 Nature paper Highly Accurate Protein Structure Prediction with AlphaFold by Jumper et al. described the architecture (Evoformer plus structure module) and demonstrated reproducibility. DeepMind subsequently released the AlphaFold Protein Structure Database in 2021, providing predicted structures for over 200 million proteins, nearly every known protein sequence.

Hassabis and Jumper shared the 2024 Nobel Prize in Chemistry with David Baker for the work. The CASP14 result is among the touchstone moments of the modern AI era and demonstrated that deep learning could solve major scientific problems that had resisted decades of conventional approaches.