CycloPepper: a machine learning platform for predicting cyclization outcomes and optimizing synthesis of therapeutic...

Area of Science:

• Medicinal Chemistry
• Computational Biology
• Biotechnology

Background:

• Cyclic peptides are promising therapeutics due to their stability and binding affinity.
• On-resin head-to-tail cyclization is a key synthetic step but is challenging.
• Selecting the correct cyclization site is crucial for successful synthesis yields.

Purpose of the Study:

• To develop a machine learning (ML) model for predicting cyclic peptide cyclization outcomes.
• To create a user-friendly platform (CycloPepper) for assessing cyclization sites.
• To accelerate the discovery and synthesis of cyclic peptide therapeutics.

Main Methods:

• Generated a standardized dataset of 306 cyclic peptides using the automated CycloBot platform.
• Developed and trained a machine learning model on this dataset.
• Validated the ML model's predictions experimentally on 74 diverse cyclic peptides.

Main Results:

• The ML model achieved an average prediction accuracy of 84% for cyclization outcomes.
• Experimental validation demonstrated an 86% prediction consistency.
• The CycloPepper platform successfully identified potential cyclization sites for disease-targeting peptides.

Conclusions:

• Machine learning-assisted synthesis significantly streamlines cyclic peptide production.
• CycloPepper provides a practical tool for researchers to assess cyclization sites.
• This approach accelerates the discovery and development of novel cyclic peptide therapeutics.