Beyond efficiency: the AI-driven shift toward safer lipid nanoparticles for gene delivery

Lipid nanoparticles (LNPs) have reshaped gene delivery, most notably through their pivotal role in mRNA vaccines. However, as LNPs move beyond vaccination toward broader therapeutic applications, maximizing delivery efficiency alone is no longer sufficient. The inherent immunogenicity of both lipid and nucleic acid components initially observed in vaccines is posing new safety challenges. Herein, we advocate a shift from efficiency-driven optimization to a principled balance between safety and efficiency. We summarize recently uncovered mechanistic insights into LNP-associated immunogenicity and review strategies to mitigate these effects through rational lipid optimization. Advances in targeting, endosomal escape, and subcellular delivery are also examined, contributing to enhance overall delivery performance. The rapid iteration and the need to integrate multidimensional design parameters call for the assistance of artificial intelligence (AI)-guided approaches, that are enabling data-driven optimization of the safety-efficiency balance. This review concludes with a discussion of current progress, limitations, and future directions of the integrated wet-dry research paradigm, outlining a roadmap for next-generation LNP platforms across diverse therapeutic applications.