The protein folding problem, understanding how a linear sequence of amino acids determines a protein's three-dimensional structure, stood as one of biology's grand challenges for over fifty years. Christian Anfinsen demonstrated in 1961 that protein structure is encoded in the amino acid sequence, but predicting structure from sequence remained computationally intractable for decades. The problem's difficulty stems from the astronomical number of possible conformations a polypeptide chain can adopt, a concept illustrated by Levinthal's paradox.
DeepMind's AlphaFold2, released in 2020, effectively solved the single-chain protein structure prediction problem, achieving experimental accuracy in the CASP14 competition. This breakthrough earned Demis Hassabis and John Jumper the 2024 Nobel Prize in Chemistry, shared with David Baker for his complementary work on computational protein design. The subsequent release of the AlphaFold Protein Structure Database, containing predicted structures for over 200 million proteins, has transformed structural biology and enabled new approaches to drug discovery, enzyme engineering, and protein design that depend on structural knowledge.
Beyond single-chain prediction, the field continues to advance on more challenging problems including protein complex structure prediction, modeling of intrinsically disordered regions, and prediction of conformational dynamics. Companies like EvolutionaryScale and academic groups are developing protein language models that capture folding principles implicitly through training on evolutionary sequence data. These models can predict the effects of mutations on stability and function, guide protein engineering campaigns, and generate novel sequences that fold into desired structures, representing a fundamentally new paradigm for understanding and manipulating protein folding.