The Nobel Prize Celebrates Breakthroughs. Who Celebrates the Foundations?
From a 1964 bacterial genetics experiment to AlphaFold, the genealogy of protein modeling is longer and more crowded than any citation can hold.
By The Editors/November 9, 2026/2 min read
In October 2024 the Nobel Committee for Chemistry split its prize in two. Half went to David Baker for computational protein design. The other half went to Demis Hassabis and John Jumper for predicting protein structures with AlphaFold. Citations have to be short. The history underneath this one runs for sixty years and passes through a great many laboratories.
Start in 1964, with bacteria and a single enzyme. Charles Yanofsky's group at Stanford was comparing mutations in the gene for tryptophan synthetase with changes in the protein it encodes. The order of the mutations along the gene matched the order of the amino acid changes along the protein. Colinearity sounds obvious today because that generation of geneticists made it obvious.
The same year, his group reported something stranger. A damaging amino acid change could be offset by a second change at a different position. Two distant points on a chain could depend on each other.
From covariation to structure
Decades later, that kind of dependence became a computational question. In 1994 a team including Chris Sander and Alfonso Valencia asked whether correlated mutations across protein families could reveal which residues touch. The signal was real and noisy, because correlation travels through intermediates: if A affects B and B affects C, A and C look linked too.
Alan Lapedes and colleagues sketched a statistical way to separate direct couplings from indirect ones in 1999. It became practical around 2009, when Martin Weigt and colleagues applied a direct-coupling method to real proteins. By 2011 studies led by Debora Marks and Faruck Morcos were recovering protein contacts and even whole folds from sequence data alone. Running alongside that work, Baker's Rosetta had been assembling structures from fragments since the 1990s. AlphaFold's performance at CASP14 in 2020 drew on all of this context, and on enormous sequence databases built by people who will never be named in Stockholm.
None of this makes the 2024 prize a mistake. Baker, Hassabis and Jumper did what their citations say they did. Nor is there a straight line from Yanofsky's bacteria to a neural network. Coevolutionary modeling came from many groups over many years, and Baker's design work has a separate history of its own.
The question worth asking is what happens to everyone else when a story gets compressed to fit on a medal.
Sources
- Yanofsky C, Carlton BC, Guest JR, Helinski DR, Henning U. On the colinearity of gene structure and protein structure. PNAS, 1964.
- Yanofsky C, Horn V, Thorpe D. Protein structure relationships revealed by mutational analysis. Science, 1964.
- Göbel U, Sander C, Schneider R, Valencia A. Correlated mutations and residue contacts in proteins. Proteins, 1994.
- Lapedes AS, Giraud BG, Liu L, Stormo GD. Correlated mutations in models of protein sequences: phylogenetic and structural effects. 1999.
- Simons KT, Kooperberg C, Huang E, Baker D. Assembly of protein tertiary structures from fragments. J Mol Biol, 1997.
- Weigt M, White RA, Szurmant H, Hoch JA, Hwa T. Identification of direct residue contacts in protein-protein interaction by message passing. PNAS, 2009.
- Marks DS et al. Protein 3D structure computed from evolutionary sequence variation. PLoS ONE, 2011.
- Morcos F et al. Direct-coupling analysis of residue coevolution captures native contacts across many protein families. PNAS, 2011.
- Jumper J et al. Highly accurate protein structure prediction with AlphaFold. Nature, 2021.
