The Biggest Challenge for Prediction of Membrane Permeability of Cyclic Peptides: Scarce Data Availability

Key Points

• This work aims to address the challenges in predicting membrane permeability of cyclic peptides due to scarce data.
• Utilized an AI model integrating a high-throughput assay
• Classified billions of cyclic peptides based on permeability-related characteristics
• Employed generative AI to design new permeable cyclic peptides
• Highlighted the challenges posed by limited ground truth biological data
• Reported the design of de novo cyclic peptides capable of higher permeability with polar groups