Trajectory-based training enables protein simulations with accurate folding and Boltzmann ensembles in cpu-hours.
John M. JumperNabil F. FarukKarl F. FreedTobin R. SosnickPublished in: PLoS Comput. Biol. (2018)
Keyphrases
- protein folding
- molecular dynamics
- protein structure prediction
- molecular dynamics simulations
- amino acid sequences
- computationally efficient
- protein sequences
- amino acids
- supervised learning
- training process
- protein structure
- neural network
- ensemble learning
- spherical harmonics
- high accuracy
- training set
- training data