Computer-Aided optimization of quinazoline-based Aurora kinase inhibitors for enhanced metabolic stability

The quinazoline-based Aurora kinase inhibitor 9h, previously developed by our group, exhibited potent efficacy both in vitro and in vivo. However, its further development was hindered by a short half-life and insufficient metabolic stability. To address these limitations, we employed machine learning-driven in silico models to predict 9h's key metabolic sites. Based on these predictions, electronic isosteric substitutions and metabolic blocking strategies were employed to optimize these sites, yielding a series of new quinazoline derivatives. Subsequent liver microsomal stability assays identified compound 4q as the lead candidate, exhibiting significantly enhanced metabolic stability with a 14-fold and 4.5-fold increase in half-life relative to 9h in rat and human microsomes, respectively. Pharmacokinetic evaluation in rats revealed 4q possesses a favorable profile, with an oral half-life of 5.5 h, an intravenous half-life of 4.1 h, and an oral bioavailability of 58 %. Importantly, 4q maintained potent inhibitory activity in vivo, achieving a tumor growth inhibition rate of 56.7 %, consistent with its robust in vitro efficacy. Molecular docking simulations further elucidated 4q's binding mode with Aurora kinase, providing a structural basis for its activity.