RoseTTAFold2

Fast, accurate structure prediction with 3-track networks

RoseTTAFold2 uses a three-track network that jointly reasons over sequences, pairwise distances and coordinates to predict monomer and complex structures. It accepts single or paired MSAs and handles flexible multidomain proteins well. Predictions include per-residue confidence estimates for filtering low-quality regions.

RoseTTAFold2 is being onboarded — request access and be first in line.