Improving De-novo Protein Design Success with a Machine Learning Model
Susan C. Kleinfelter
Tr-Rosetta, a machine learning model, was used with Rosetta fastdesign for focused redesign of problem areas in 4000 de-novo protein designs by players of the citizen science game Foldit. Tr-Rosetta improved fragment quality, a predictor of design success, allowing “rescue” of designs that would otherwise not fold as intended.