RoseTTAFold
Era 4 · Deep learning and foundation models (2012–2021) › AI for science
- Lineage
- Acting in the world
- Kind
- Publication
Who
University of Washington
What happened
On July 15, 2021, Minkyung Baek, David Baker, and colleagues at the University of Washington's Institute for Protein Design publish RoseTTAFold in Science: a three-track neural network that predicts protein structure with accuracy close to that of AlphaFold 2.
Why it mattered
A second method of comparable accuracy, open and able to run on a gaming computer, which also predicts how several proteins bind together.
The full story
In late 2020 AlphaFold 2's results at CASP14 showed that protein structure could be predicted far more accurately than anyone had managed before, but DeepMind had not yet published how it did it. David Baker's lab in Seattle set out to reach that performance on its own.
The result was RoseTTAFold, a three-track network. One track analyzes patterns in the protein's sequence; another, how its amino acids interact; and the third, its possible three-dimensional structure. Information flows back and forth between the tracks as the network computes. According to the University of Washington, the team largely recreated the performance DeepMind had shown at CASP14, and the program could compute a structure in as little as ten minutes on a single gaming computer.
The paper, with 32 authors, came out online on July 15, 2021, the same day Nature published AlphaFold 2. Its title also announced the prediction of protein-protein interactions. The team released the program openly, and more than 140 research groups downloaded it within a short time.
In 2024 Baker received half of the Nobel Prize in Chemistry, for computational protein design, and Hassabis and Jumper shared the other half.
Builds on
AlphaFold 2 (November 2020)
Independent convergence
Baker's lab reached, on its own, performance close to what DeepMind had shown at CASP14, before the method was published; the two papers came out on the same day
Sources
Original source
- Accurate prediction of protein structures and interactions using a three-track neural network. Science, 373(6557), pp. 871-876 (opens in a new tab). Minkyung Baek, Frank DiMaio, Ivan Anishchenko, et al., with David Baker, 2021
Scholarship and context
How to cite
“RoseTTAFold.” The History of AI, revised November 13, 2026. https://thehistoryofai.com/en/milestone/rosettafold
This entry was last reviewed on November 13, 2026.