Peptide-aware chemical language model successfully predicts membrane diffusion of cyclic peptides

Area of Science:

• Computational chemistry
• Biophysics
• Pharmacology

Background:

• Language models have improved biological data analysis, particularly for small molecules and natural peptides.
• Predicting membrane diffusion for modified peptides remains a significant challenge in drug development.

Purpose of the Study:

• Introduce PeptideCLM, a chemical language model specifically designed for peptides.
• Enhance the prediction of membrane diffusion for peptides, including those with modifications.

Main Methods:

• Developed PeptideCLM, a peptide-focused chemical language model.
• Encoded peptides with modifications, unnatural amino acids, and cyclizations.
• Assessed model performance by predicting cyclic peptide membrane diffusion.

Main Results:

• PeptideCLM demonstrated superior predictive power for cyclic peptide membrane diffusion compared to existing models.
• The model effectively encodes complex peptide features like modifications and cyclizations.

Conclusions:

• PeptideCLM offers a versatile and adaptable solution for predicting peptide membrane diffusion.
• The model's ability to handle macromolecules via chemical string notation opens new research avenues.