Deep Learning Enabled Design of Functional Cyclic Peptides

Computational protein design has undergone a paradigm shift since the release of deep learning models such as AlphaFold2. Fast, accurate and accessible design tools for numerous applications have been developed at accelerated rates for protein design of all flavors. Cyclic peptide drug design however, often relies on non-canonical linkages and building blocks that these tools were incapable of utilizing in their infancy. In this seminar, I will describe the development and experimental validation of AfCycDesign and RFpeptides, for atomically accurate, high-affinity cyclic peptide binder design, using deep learning models.