Structure-Based Prediction of Molecular Interactions for Stabilizing Volatile Drugs

Background/Objectives: The high volatility of volatile drugs significantly restricts their clinical applicability. Although excipients capable of strong interactions can reduce volatilization, conventional screening methods rely on empirical trial-and-error, resulting in low efficiency and high resource consumption. To address this limitation, this study introduces an artificial intelligence (AI)-driven strategy for screening drug–excipient interactions. Using d-borneol as a model drug, this approach aims to efficiently identify strongly interacting excipients and develop stable nano-formulations. Methods: High-throughput simulations were performed using the Protenix structure prediction model to evaluate interactions between d-borneol and 472 FDA-approved excipients. The top 50 candidate excipients were selected based on these simu-lations. Molecular docking and stability experiments were conducted to validate the predictions. Results: Molecular docking and stability experiments confirmed the consistency between predicted and experimental results, validating the model’s reliability. Among the candidates, soybean phospholipid (PC) was identified as the optimal excipient. A lyophilized liposomal formulation prepared with PC significantly suppressed the volatilization of d-borneol and improved both thermal and storage stability. Mechanistic investigations indicated that d-borneol stably incorporates into the hydro-phobic region of phospholipids, enhancing membrane ordering via hydrophobic interactions without disturbing the polar headgroups. Conclusions: This study represents the first application of a structure prediction model to excipient screening for volatile drugs. It successfully addresses the stability challenges associated with d-borneol and offers a new paradigm for developing nano-formulations for volatile pharmaceuticals.