Discovery of New Pentapeptide Inhibitors Against Amyloid-β Aggregation Using Word2Vec and Molecular Simulation

Alzheimer's disease (AD) is characterized by the aggregation of amyloid-β (Aβ) peptides into toxic oligomers and fibrils. The efficacy of existing peptide inhibitors based on the central hydrophobic core (CHC) sequence of Aβ42 remains limited due to self-aggregation or poor inhibition. This study aimed to identify novel pentapeptide inhibitors with high similarity and low binding energy to the CHC region LVFFA using a new computational screening workflow based on Word2Vec and molecular simulation. The antimicrobial peptides and human brain protein sequences were used for training the Word2Vec model. After tuning the parameters of the Word2Vec model, 1017 pentapeptides with high similarity to LVFFA were identified. Molecular docking was employed to estimate the affinity of the pentapeptides for the target of Aβ14-42 pentamer, and 103 peptides with favorable docking scores were obtained. Finally, five pentapeptides with a low binding energy and high binding stability via molecular dynamics simulation were experimentally validated using thioflavin T assays. Surprisingly, one pentapeptide, i.e., PALIR, exhibited significant inhibition surpassing the positive control LPFFN. This study demonstrates an effective combinatorial strategy to discover new peptide inhibitors. With PALIR representing a promising lead candidate, further optimization of PALIR could aid in the development of improved therapies to prevent amyloid toxicity in AD.