Identification of a novel Aurora B inhibitor using the AI-driven drug screening and docking-based traditional screening

Aurora B, a subtype of Aurora kinases that functions as a serine/threonine kinase, playing a vital role in the process of mitosis, is often overexpressed in certain tumor cells leading to tumorigenesis and progression. Therefore, the development of small molecule inhibitors targeting Aurora B holds promise for providing new options for some cancer patients. In this study, we efficiently screened 4 compounds from MCE compound database using a combination of machine learning-based screening and structure-based screening. The results showed that 2 compounds exhibited strong Aurora B inhibitory activity in a homogeneous time-resolved fluorescence (HTRF) assay, indicating a high hit rate for this screening method. Among them, compound 4 demonstrated optimal inhibitory activity against Aurora B, with an IC₅₀ value of 15.54 nM, comparable to Aurora B inhibitors that have entered clinical trials. In vitro experiments indicated that compound 4 effectively inhibited Huh-7 and Huh-6 cells, with IC₅₀ values of 0.9 μM and 1.8 μM, respectively. Molecular dynamics simulation results revealed that the compound binds to the ATP binding pocket of Aurora B, forming hydrogen bond interactions with Glu171 and Glu220, salt bridges with Asp234 and Glu177, and a pi-cation interaction with Arg97. In summary, by integrating multi-modal screening approaches, we successfully identified a potent Aurora B inhibitor with in vitro antitumor activity, providing lead compounds for subsequent drug development.