AlphaFold: influences, peers and legacy

Every connection with its receipt

AlphaFold sits at the junction of two long lineages: the 50-year structural-biology quest launched by Anfinsen and Levinthal, and the deep-learning revolution that DeepMind had already proven on Go. This mix traces the crystallographers, coevolution statisticians and game programmers whose work it absorbed, plus the open databases and successor models it set loose.

The Kynda mix for AlphaFold

• Key Influence · Principles that Govern the Folding of Protein Chains by Christian B. Anfinsen (1973). Anfinsen's ribonuclease refolding experiments established that a protein's amino-acid sequence alone encodes its three-dimensional structure — the thermodynamic hypothesis that makes sequence-to-structure prediction logically possible in the first place. Every CASP round, and AlphaFold's entire premise of reading a folded shape out of a sequence string, rests on this claim. DeepMind's Nature papers frame the problem exactly in these terms.
• Influencia Obscura · How to Fold Graciously by Cyrus Levinthal (1969). Levinthal's short Mössbauer-symposium paper posed the paradox that a chain cannot sample all conformations in biological time, so folding must follow pathways — the argument that reframed folding as a computational rather than brute-force search problem. It is rarely read and constantly cited, and it defines precisely the intractability that a learned prior, rather than physical simulation, sidesteps.
• Local Roots · Structure of Haemoglobin by Max Perutz (1960). Perutz's decades-long haemoglobin work at the MRC Laboratory of Molecular Biology in Cambridge, alongside Kendrew's myoglobin, created the experimental structures that eventually filled the Protein Data Bank — AlphaFold's training data. British structural biology's Cambridge lineage is the direct local ancestry of a London-headquartered lab solving the same problem by other means half a century later.
• Beyond the Medium · Theme Park by Bullfrog Productions (1994). Hassabis co-designed and programmed AI for Bullfrog's Theme Park as a teenager under Peter Molyneux before founding Elixir Studios and later DeepMind. The habit of building adaptive simulated systems — and of treating hard problems as games with scoreboards — shapes DeepMind's method, from benchmark-driven Go and Atari work to entering CASP as a competition to be won outright.
• Peer · RoseTTAFold by David Baker (2021). Baker's Institute for Protein Design team, having watched AlphaFold 2's CASP14 result, reverse-engineered the key idea into a three-track network reasoning jointly over sequence, distances and coordinates, publishing in Science days before AlphaFold's Nature paper. Baker's Rosetta had been the dominant prediction software for twenty years; the two labs are the field's defining rivalry and mutual spur.
• Key Collaborator · Critical Assessment of Structure Prediction (CASP) by John Moult (1994). Moult founded the biennial blind-prediction experiment in 1994, supplying unreleased experimental structures as held-out test data and the GDT scoring metric by which AlphaFold's 2018 and 2020 wins were measured. Without this rigorously adversarial external benchmark, DeepMind's claim would have been unfalsifiable; Moult's own verdict that the problem was in some sense solved is what gave the result its weight.
• Legacy · Boltz-1 by MIT Jameel Clinic (2024). An openly licensed, fully open-source reimplementation of biomolecular complex prediction at AlphaFold 3 level, released explicitly in response to DeepMind's decision to withhold AlphaFold 3 code. Its existence is a direct argument with its predecessor about what open science requires, and its architecture openly builds on the Evoformer-and-diffusion lineage AlphaFold established.

What influenced AlphaFold

• Theme Park by Bullfrog Productions (1994). Theme Park (Bullfrog Productions) — culture for AlphaFold “He began by playtesting on Syndicate and then at 17 co-designing and lead-programming on the 1994 game Theme Park, with the game's designer Peter Molyneux.” (en.wikipedia.org)
• Gödel, Escher, Bach: An Eternal Golden Braid by Douglas Hofstadter (1979). Gödel, Escher, Bach: An Eternal Golden Braid (Douglas Hofstadter) — culture for AlphaFold (wikidata.org)
• Principles that Govern the Folding of Protein Chains by Christian B. Anfinsen (1973). Principles that Govern the Folding of Protein Chains (Christian B. Anfinsen) — titan for AlphaFold
• Foldit by Seth Cooper (2008). Foldit (Seth Cooper) — ghost for AlphaFold
• ImageNet Classification with Deep Convolutional Neural Networks by Alex Krizhevsky (2012). ImageNet Classification with Deep Convolutional Neural Networks (Alex Krizhevsky) — titan for AlphaFold
• How to Fold Graciously by Cyrus Levinthal (1969). How to Fold Graciously (Cyrus Levinthal) — ghost for AlphaFold
• Protein 3D Structure Computed from Evolutionary Sequence Variation by Debora Marks (2011). Protein 3D Structure Computed from Evolutionary Sequence Variation (Debora Marks) — ghost for AlphaFold
• The Structure and Action of Proteins by Irving Geis (1969). The Structure and Action of Proteins (Irving Geis) — culture for AlphaFold
• Attention Is All You Need by Ashish Vaswani (2017). Attention Is All You Need (Ashish Vaswani) — titan for AlphaFold

Peers and kindred spirits

• AlphaFold Protein Structure Database by EMBL-EBI (2021). AlphaFold Protein Structure Database (EMBL-EBI) — collaborator for AlphaFold “== Database of protein models generated by AlphaFold == The AlphaFold Protein Structure Database (AFDB), a joint project between AlphaFold and EMBL-EBI, was launched on July 22, 2021.” (en.wikipedia.org)
• RoseTTAFold by David Baker (2021). RoseTTAFold (David Baker) — peer for AlphaFold “Demis Hassabis and John Jumper shared one half of the 2024 Nobel Prize in Chemistry, awarded "for protein structure prediction," while the other half went to David Baker "for computational protein design." Hassabis and Jumper had previou…” (en.wikipedia.org)
• Critical Assessment of Structure Prediction (CASP) by John Moult (1994). Critical Assessment of Structure Prediction (CASP) (John Moult) — collaborator for AlphaFold
• Language Models are Few-Shot Learners by OpenAI (2020). Language Models are Few-Shot Learners (OpenAI) — peer for AlphaFold
• Molecular Structure of Nucleic Acids by James Watson (1953). Molecular Structure of Nucleic Acids (James Watson) — geography for AlphaFold
• ESMFold by Meta AI (2022). ESMFold (Meta AI) — peer for AlphaFold
• Structure of Haemoglobin by Max Perutz (1960). Structure of Haemoglobin (Max Perutz) — geography for AlphaFold
• The Protein Data Bank by Helen M. Berman (2000). The Protein Data Bank (Helen M. Berman) — collaborator for AlphaFold
• Computing Machinery and Intelligence by Alan Turing (1950). Computing Machinery and Intelligence (Alan Turing) — geography for AlphaFold

Who AlphaFold influenced

• Boltz-1 by MIT Jameel Clinic (2024). Boltz-1 (MIT Jameel Clinic) — legacy for AlphaFold
• Chai-1 by Chai Discovery (2024). Chai-1 (Chai Discovery) — legacy for AlphaFold