Succinct Amyloid and Nonamyloid Patterns in Hexapeptides.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology and Bioinformatics
• Artificial Intelligence in Science

Background:

• Hexapeptides serve as crucial models for studying polypeptide and protein amyloid formation.
• Publicly available experimental databases now offer extensive amyloidogenic labels for peptides.
• Artificial intelligence (AI) offers a powerful approach for predicting peptide amyloid states.

Purpose of the Study:

• To apply the Budapest Amyloid Predictor (BAP) for discovering novel amyloidogenic and nonamyloidogenic hexapeptide patterns.
• To establish succinct rules for understanding peptide amyloid formation using AI.
• To explore the impact of restricted amino acid substitutions on amyloidogenicity prediction.

Main Methods:

• Utilized a Support Vector Machine (SVM)-based AI classifier, the Budapest Amyloid Predictor (BAP).
• Trained and tested the AI model on large experimental databases of amyloidogenic labels.
• Analyzed specific hexapeptide patterns, including those with variable positions ('x'), to identify amyloidogenic and nonamyloidogenic sequences.

Main Results:

• Achieved prediction accuracy between 80% and 84% for identifying amyloidogenic and nonamyloidogenic hexapeptide patterns.
• Discovered succinct patterns, such as CxFLWx (amyloidogenic) and PxDxxx (nonamyloidogenic), offering novel insights.
• Demonstrated that restricted substitutions, particularly with hydrophobic amino acids, can yield highly amyloidogenic patterns (e.g., xxxFxx).

Conclusions:

• AI-driven pattern discovery provides valuable, succinct rules for predicting peptide amyloidogenicity.
• The identified patterns significantly enhance the understanding of factors governing amyloid formation in peptides.
• This work represents a novel application of AI in discovering sequence-based rules for amyloid states.