ANuPP: A Versatile Tool to Predict Aggregation Nucleating Regions in Peptides and Proteins

Area of Science:

• Biochemistry
• Computational Biology
• Structural Biology

Background:

• Protein aggregation is implicated in various diseases.
• Current methods for identifying aggregation-prone regions (APRs) primarily use residue-level features.
• Understanding the initiation of protein aggregation requires exploring finer-level characteristics.

Purpose of the Study:

• To investigate the importance of atomic-level characteristics in protein and peptide aggregation.
• To develop a computational tool for predicting aggregation-nucleating regions using atomic features.
• To enhance the accuracy of identifying aggregation-prone regions (APRs) compared to existing methods.

Main Methods:

• Developed an ensemble classifier named ANuPP (Aggregation Nucleation Prediction Pipeline).
• Utilized atomic-level features, including functional group characteristics, for prediction.
• Evaluated ANuPP using 10-fold cross-validation on a large hexapeptide dataset and a blind test set.
• Assessed performance on identifying APRs in a set of 37 proteins.

Main Results:

• ANuPP achieved an AUC of 0.831 with 77% accuracy on 10-fold cross-validation.
• In a blind test, ANuPP reached an AUC of 0.883 with 83% accuracy.
• Demonstrated an average SOV of 48.7% in identifying APRs across 37 proteins.
• ANuPP outperformed existing methods in both hexapeptide prediction and APR identification.

Conclusions:

• Atomic-level characteristics are crucial for understanding and predicting protein aggregation initiation.
• ANuPP provides a more accurate approach to identifying aggregation-nucleating regions.
• The findings highlight the diversity of atomic-level origins and mechanisms in protein aggregation.
• A web server for ANuPP is available for public use.