EnsembleCycPerm: Interpretable Modeling of Cyclic Peptide Permeability through Solvent-Dependent Conformational...

Area of Science:

• Computational chemistry
• Medicinal chemistry
• Pharmacology

Background:

• Cyclic peptides show therapeutic promise but face challenges in clinical translation due to poor membrane permeability.
• This limitation stems from complex, solvent-dependent conformational ensembles.
• Existing models often struggle to capture these dynamic behaviors.

Purpose of the Study:

• To develop a robust modeling framework for predicting cyclic peptide membrane permeability.
• To integrate sequence and structural information, explicitly representing solvent-dependent conformational ensembles.
• To establish a generalizable approach for cyclic peptide design and ADMET prediction.

Main Methods:

• A novel modeling framework inspired by the 'molecular chameleon' concept was developed.
• The model explicitly represents solvent-dependent conformational ensembles in aqueous and nonpolar environments.
• Performance was evaluated using random and scaffold split strategies on the CycPeptMPDB dataset.

Main Results:

• The model demonstrated strong predictive performance (MAE = 0.29, R = 0.85, R² = 0.70) under random split.
• Robust performance was maintained under scaffold split evaluation (MAE = 0.34, Pearson R = 0.68).
• Analysis revealed that polarity shielding, quantified by ΔPSA3D, is a key feature associated with permeability.

Conclusions:

• The developed model accurately predicts cyclic peptide permeability, addressing a key challenge in their therapeutic application.
• Permeability is shown to arise from ensemble reorganization rather than a single conformational change.
• This framework offers a generalizable tool for cyclic peptide lead optimization and ADMET prediction.