Abstract by Jing Zhao

Protein–protein interactions (PPIs) are central to cellular regulation but remain difficult to target with conventional small molecules. Molecular glues, which induce or stabilize PPIs, offer a promising strategy to reprogram cellular networks, yet their discovery has largely been serendipitous and lacks general design principles. Macrocyclic scaffolds are a common feature of several natural molecular glues, suggesting that cyclic peptides may be well suited for this function. The Random nonstandard Peptides Integrated Discovery (RaPID) system enables the generation of diverse, high-affinity macrocyclic peptides and represents a powerful platform for exploring molecular glue discovery.

In this thesis, I combine deep mutational scanning (DMS) with RaPID-based selection strategies to investigate how RaPID-derived cyclic peptides engage protein targets, to evaluate their tunability for associating two proteins, and to assess whether RaPID can be adapted into a systematic source of molecular glues. DMS analysis of cyclic peptides targeting hnRNPA1 RRM2, human ASPA, and the bacterial β-clamp identifies residues essential for binding, and reveals that certain RaPID-derived peptide can induce β-clamp dimerization, functioning as a molecular glue.

To enable rational glue discovery, I develop alternating and fixed-template RaPID strategies to directly select for peptides that associate two proteins. Using these approaches, I identify cyclic peptides that form ternary complexes, including a bifunctional molecule that associates FRB domain of mTOR and FKBP12, and a molecular glue that induces proximity between Cyclophilin A and PCNA.

Overall, this work establishes RaPID as a versatile platform for discovering both molecular glues and bifunctional compounds, providing a foundation for expanding the therapeutic potential of this emerging drug modality.