In Silico Screening of Small Molecule Inhibitors for Amyloid-β Aggregation

The self-aggregation of amyloid-β (Aβ) into fibrils is a hallmark of Alzheimer's disease (AD). Inhibition of Aβ aggregation with small molecule compounds represents a promising therapeutic strategy for AD. However, designing effective ligands is challenging due to the lack of structural information about the binding target. In this work, ligand-based virtual screening for Aβ aggregation inhibitors was conducted using molecular representation models. After comparing the performance of four machine learning models and two molecular representation models using inhibitor data from the literature, the optimal model was applied for prediction, and novel small molecules were screened out. Among the top-ranked compounds, 11 were experimentally investigated and four of them demonstrated better inhibition effects than EGCG. The interactions between the four candidates and Aβ aggregates were analyzed via molecular dynamics simulation, revealing that the binding of a small molecule compound brought interference to the fluctuation mode of the pentamer and weakened its stability. This work highlights the efficacy of molecular representation based virtual screening approach for discovering Aβ aggregation inhibitors with novel structure and high affinity. The experimental hit rate using this virtual screening approach was 36.4%.