Combining AI design and molecular dynamics to disrupt amyloid aggregation in Alzheimer’s disease

Self-aggregation of amyloid beta (Aβ) peptides in neural cells results in nerve cell toxicity and is one of the indications of Alzheimer’s disease. Therapeutic approaches targeting the early stages of Aβ peptide aggregation offer a promising strategy to slow down or prevent peptide oligomerization and plaque formation. In this study, de novo peptides of three varying lengths (15, 20, and 42 amino acids) were generated using the RoseTTAFold diffusion model with high predicted binding affinity to the 42-amino acid Aβ peptide of H. sapiens. Molecular docking of the selected designed peptides and the H. sapiens Aβ peptide demonstrated that the designed 20- and 42-amino acid peptides interact with the H. sapiens Aβ with a comparable docking score and electrostatic energy, forming various interactions with H. sapiens Aβ. Extensive molecular dynamics simulations showed that the 15- and 20-amino-acid peptides formed significantly more stable complexes with H. sapiens Aβ, resulting in less conformational flexibility and disruption of aggregation-prone regions. Whereas, only moderate stabilization of Aβ and limited inhibitory effects could be observed for the 42-residue peptide. Ultimately, the results suggest that both 15- and 20-amino-acid peptides can strongly inhibit the structural transitions associated with early oligomerization, and thus hold great promise as modulators of downstream fibril and plaque formation. These findings highlight the potential of computationally designed peptides with AI-powered approaches as innovative and practical therapeutic options for neurodegenerative diseases.