Universal design of peptide binders for therapeutic and diagnostic modalities
Abstract
Structure prediction has revolutionised protein design, yet the de novo generation of functional modulators, specifically small peptides that actively alter protein states, remains a grand challenge. Short peptides offer distinct therapeutic advantages, including membrane permeability and reduced steric hindrance, making them ideal candidates for targeting "undruggable" interfaces.
We previously established EvoBind, a validated framework for peptide binder design (https://www.nature.com/articles/s42004-025-01601-3). Building on this foundation, we are now transitioning from static binder design to dynamic functional engineering. This continuation focuses on three major advancements:
1) Chemical Expansion: We have deployed RareFold and EvoBindRare, extending our design capabilities to non-canonical amino acids (NCAAs) to access novel chemical space and improve proteolytic stability.
2) Functional Agonism: We are applying EvoBind and RareFoldGPCR to design de novo agonists for G-protein coupled receptors (e.g., GLP1R, GCGR), a task that requires capturing subtle conformational shifts unavailable to standard folding models.
3)Multi-body Logic: We are now introducing the development of EvoBind-multimer, a new framework for designing molecular glues and ternary complexes. This moves beyond binary interactions to engineer cooperative binding and chemically induced proximity.
These tasks require computationally intensive modelling of conformational ensembles and multi-state physics, necessitating high-performance computing for massive decoy sampling and transfer learning on sparse datasets. We aim to apply this precision peptide technology to therapeutic targets in cancer, diabetes, and viral entry inhibition.
