RoseTTAFold predicts how a chain of amino acids folds into a three-dimensional protein. Knowing a protein’s shape helps scientists understand what it does and how to target it with drugs. The model was developed at the Institute for Protein Design and the University of Washington, around the same time as AlphaFold, and offered a comparable approach with a three-track neural network.
The work has grown into a family. RoseTTAFold All-Atom models complexes that include small molecules, DNA, and other components, and RFdiffusion generates new protein backbones for design tasks such as binders and enzymes. Together these tools support a shift from predicting existing proteins to creating new ones, with applications in medicine and materials.
The code and models are available for research use through GitHub and a web server, with licensing details that vary by release. Predictions need experimental validation, and running the models takes computing resources and expertise. Alternatives include AlphaFold, ESMFold, OpenFold, Boltz, and Chai. RoseTTAFold suits structural biologists and protein designers.