Amyloidogenic motifs revealed by n-gram analysis

Area of Science:

• Biochemistry
• Computational Biology
• Protein Science

Background:

• Amyloid proteins are linked to neurodegenerative diseases such as Alzheimer's and Creutzfeldt-Jakob's.
• Protein aggregation is initiated by specific segments known as "hot spots."

Purpose of the Study:

• To identify sequence patterns defining amyloidogenic hot spots.
• To develop a predictive tool for amyloidogenicity.

Main Methods:

• Trained predictors using n-grams and random forest classifiers.
• Tested 524,284 reduced amino acid alphabets to optimize prediction.
• Developed the AmyloGram predictor based on the best-performing alphabet.

Main Results:

• AmyloGram achieved high performance metrics (AUC: 0.90, MCC: 0.63) on an external dataset.
• Identified sequential patterns correlated with hydrophobicity, beta-sheet propensity, and low flexibility.
• Discovered 15 informative n-grams experimentally validated.

Conclusions:

• AmyloGram accurately predicts amyloidogenic regions.
• The findings enhance understanding of amyloid formation mechanisms.
• AmyloGram is available as a web server and R package for research use.