AI-Driven Development of Next Generation Lipid Nanoparticles for mRNA Delivery

The clinical success of mRNA vaccines has catalyzed rapid innovation in lipid nanoparticle (LNP) delivery systems. However, rationally engineering LNPs for extrahepatic tissues and emerging RNA modalities remains a major bottleneck due to the vast chemical space and lack of predictive design rules. Here, we present two integrated platforms, AGILE (AI-Guided Ionizable Lipid Engineering) and LUMI-lab (Lipid Utility Modeling and Innovation), that leverage deep learning and autonomous high-throughput experimentation to accelerate LNP discovery for mRNA therapeutics.
AGILE employs molecular graph-based learning and structure-aware pretraining to predict LNP performance across diverse lipid scaffolds, enabling the identification of novel ionizable lipids with enhanced mRNA delivery to non-hepatic tissues. Building on this, LUMI-lab integrates generative modeling with automated synthesis and screening to iteratively design, test, and optimize high-performing lipids in a closed-loop system. The foundation of LUMI-lab, LUMI-model, was pretrained on over 28 million molecules using 3D coordinate-based contrastive learning, capturing nuanced structural features that inform delivery efficacy. Using these platforms, we discovered new lipids that demonstrated superior mRNA transfection efficiency and CRISPR-based gene editing activity in vitro and in vivo. Collectively, our work establishes a scalable AI-guided framework for next-generation LNP design, offering a path toward programmable RNA delivery vehicles tailored for specific tissues, disease contexts, and therapeutic applications.