Design of permeability-optimized target-binding macrocycles via direct preference optimization

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Drug Discovery

Background:

• Macrocyclic peptides show therapeutic promise for difficult targets like protein-protein interactions.
• Clinical application is hindered by poor membrane permeability, limiting oral bioavailability and intracellular access.
• Current design methods often fail to optimize permeability alongside binding affinity and structural validity.

Purpose of the Study:

• To introduce CycDiff-DPO, a novel diffusion framework for designing macrocyclic peptides with enhanced membrane permeability.
• To integrate a Caco-2 permeability predictor and preference learning into the generative process.
• To enable co-optimization of target binding and membrane permeability in peptide design.

Main Methods:

• Developed CycDiff-DPO, a preference-aligned diffusion model for macrocyclic peptide generation.
• Utilized a Caco-2 permeability predictor to rank and filter generated peptide candidates.
• Constructed preference pairs to guide the generative distribution towards permeability-favorable chemical space.
• Benchmarked performance across 56 protein targets against existing structure-based design methods.

Main Results:

• CycDiff-DPO demonstrated improved predicted Caco-2 and PAMPA permeability across multiple independent predictors.
• Designs exhibited superior binding energetics compared to baseline methods while maintaining stereochemical quality.
• Case studies on Keap1-Nrf2 and SPSB2-iNOS showed designs recapitulating key interactions and stable molecular dynamics.
• The framework successfully balanced target binding competence with enhanced permeability.

Conclusions:

• CycDiff-DPO offers a robust framework for designing macrocyclic peptides with improved membrane permeability.
• This approach facilitates the development of peptide therapeutics with enhanced bioavailability and intracellular target access.
• The method holds broad potential for therapeutic applications targeting challenging diseases.