A groundbreaking AI-powered platform has dramatically accelerated the discovery of new materials for delivering mRNA into human cells. The system, called LUMI-lab (Large-scale Unsupervised Modeling followed by Iterative experiments), combines artificial intelligence with robotics to create a self-driving lab that autonomously designs, synthesizes, and tests new lipid nanoparticles (LNPs) for mRNA delivery. Developed by researchers at the University of Toronto’s Leslie Dan Faculty of Pharmacy, the platform has identified brominated lipids—previously unexplored in mRNA delivery—as a major factor in enhancing transfection efficiency. Published in Cell, the study details how LUMI-lab, supported by an AC Translation research grant from U of T’s Acceleration Consortium, integrates a molecular foundation model with automated experimental workflows. The system was pretrained on over 28 million molecular structures, enabling it to learn broad chemical patterns before focusing on specific delivery challenges. This foundation allows the AI to make intelligent predictions even in data-scarce fields like mRNA therapeutics. Over ten active-learning cycles, LUMI-lab designed and tested more than 1,700 novel lipid nanoparticles. Remarkably, it independently discovered that brominated lipid tails significantly boost mRNA delivery into human lung cells—outperforming even the lipid used in Moderna’s approved COVID-19 vaccine. The discovery was unexpected, as bromination had not been previously considered a key design feature in LNPs. Despite making up only 8% of the initial compound library, brominated lipids accounted for over half of the top-performing candidates. “This is a major leap forward,” said Bowen Li, GSK Chair in Pharmaceutics and Drug Delivery at the Leslie Dan Faculty of Pharmacy and affiliate scientist at the Princess Margaret Cancer Centre, University Health Network. “The AI didn’t need to be told to look for bromination. It identified it as a meaningful design element on its own.” Currently, only three lipid nanoparticles have received FDA approval for mRNA delivery, limiting the scope of therapeutic applications. LUMI-lab’s approach expands the design space by rapidly exploring chemical variations that would be impractical for traditional trial-and-error methods. The platform’s closed-loop system—where AI predictions feed into robotic synthesis and testing, which then refine the model—continuously improves accuracy and efficiency. Preliminary preclinical results show that the best-performing brominated lipids maintain safety profiles comparable to existing clinical-grade lipids, supporting their potential for future development. Looking ahead, Li’s team aims to evolve LUMI-lab to optimize multiple properties simultaneously—delivery potency, safety, tolerability, and tissue selectivity—rather than focusing on a single metric. By closing the loop between AI-driven insights and automated experimentation, the researchers hope to drastically shorten the timeline for developing next-generation mRNA delivery systems, unlocking new possibilities for treating cancer, genetic disorders, and infectious diseases.