AmyloPick: A New Feature Selection Method and Proper Evaluation for Amyloid Hexapeptides and Aggregation-Prone...

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Neuroscience

Background:

• Protein aggregation is critical in neurodegenerative diseases, necessitating accurate prediction tools.
• Hexapeptides serve as models for studying amyloid formation and identifying aggregation-prone regions.
• Existing computational methods for amyloid prediction have limitations in accuracy and interpretability.

Purpose of the Study:

• To develop a novel, interpretable, and accurate computational model for predicting amyloid formation in hexapeptides.
• To extend the model for identifying aggregation-prone regions (APRs) in full-length proteins.
• To establish a rigorous statistical methodology for performance assessment and comparison of prediction tools.

Main Methods:

• Designed a novel feature selection method for hexapeptides, resulting in a four-feature representation (AmyloPick model).
• Developed a classifier based on the AmyloPick representation.
• Extended the classifier to detect aggregation-prone regions (APRs) in proteins (APR-AmyloPick).
• Employed a robust statistical methodology for performance evaluation and comparison with existing tools like BAP and ANuPP.

Main Results:

• The AmyloPick classifier significantly outperformed state-of-the-art methods in hexapeptide classification, showing higher accuracy and adjusted geometric mean (AGM).
• APR-AmyloPick demonstrated comparable performance to ANuPP for APR identification, with significant outperformance in the ${\mathrm{SOV}}_{\text{Non}-\mathrm{APR}}$ metric.
• The developed model provides an interpretable four-feature representation.

Conclusions:

• The AmyloPick model offers a significant advancement in predicting protein aggregation and identifying aggregation-prone regions.
• The rigorous statistical methodology ensures reliable performance assessment and facilitates method comparability.
• A web server for the AmyloPick classifier has been developed, enhancing accessibility for researchers.