Leveraging structure-informed machine learning for fast steric zipper propensity prediction across whole proteomes

Area of Science:

• Biophysics
• Computational Biology
• Structural Biology

Background:

• Predicting amyloid fold and peptide structures is challenging.
• Recent advances include structure-based prediction and machine learning for predictive models.

Purpose of the Study:

• Develop a rapid, proteome-wide method for assessing steric zipper propensity.
• Leverage a decade of steric zipper predictions to build a machine learning model.

Main Methods:

• Utilized four million steric zipper predictions to train a machine learning model.
• Applied the model for rapid prediction of steric zipper propensity.
• Assessed zipper profiles at both protein and proteome levels.

Main Results:

• Developed a machine learning model for fast steric zipper propensity prediction.
• Identified enrichment of zipper-forming segments in yeast cell wall reorganization proteins.
• Demonstrated the model's utility for exploring amyloid formation across diverse organisms.

Conclusions:

• The developed model enables rapid, large-scale assessment of amyloidogenic sequences.
• Highlights potential roles of steric zippers in yeast cell wall organization.
• Provides a tool for evaluating zipper density in novel and designed sequences.