GAP: towards almost 100 percent prediction for β-strand-mediated aggregating peptides with distinct morphologies

Motivation: Distinguishing between amyloid fibril-forming and amorphous beta-aggregating aggregation-prone regions (APRs) in proteins and peptides is crucial for designing novel biomaterials and improved aggregation inhibitors for biotechnological and therapeutic purposes. Results: Adjacent and alternate position residue pairs in hexapeptides show distinct preferences for occurrence in amyloid fibrils and amorphous beta-aggregates. These observations were converted into energy potentials that were, in turn, machine learned. The resulting tool, called Generalized Aggregation Proneness (GAP), could successfully distinguish between amyloid fibril-forming and amorphous beta-aggregating hexapeptides with almost 100 percent accuracies in validation tests performed using non-redundant datasets. Conclusion: Accuracies of the predictions made by GAP are significantly improved compared with other methods capable of predicting either general beta-aggregation or amyloid fibril-forming APRs. This work demonstrates that amino acid side chains play important roles in determining the morphological fate of beta-mediated aggregates formed by short peptides.