AI Platform Autonomously Discovers mRNA Delivery Lipids
A self-driving AI platform called LUMI-lab has autonomously discovered novel ionizable lipids for mRNA delivery. The system, which integrates AI with robotics, screened over 1,700 candidates in a closed loop, ultimately identifying materials that achieved 20.3% gene editing efficiency in lung cells.
- The AI "brain" of the system, LUMI-model, is a foundation model pretrained on 28 million 3D molecular structures, which allows it to learn efficiently from the sparse data generated during the experimental cycles. - A key discovery was the autonomous identification of brominated lipid tails as a novel structural feature that enhances mRNA delivery, an insight not previously recognized in expert-driven lipid nanoparticle design. - The top-performing lipids discovered by the platform, including one named LUMI-6, demonstrated superior mRNA transfection efficiency in human bronchial epithelial cells compared to clinically approved benchmarks like DLin-MC3-DMA (MC3). - The 20.3% gene editing efficiency was achieved *in vivo* in murine models, delivering CRISPR-Cas9 machinery via inhalation, which reportedly surpasses the previous highest efficiency for this delivery method. - This research was led by Bowen Li at the University of Toronto's Leslie Dan Faculty of Pharmacy, and a patent has been filed for the LUMI-lab platform and the ionizable lipids it discovered. - The closed-loop workflow automates the entire design-make-test-analyze cycle, using robotic liquid handlers for synthesis and plate readers to measure mRNA transfection potency, completing ten iterative cycles to generate its findings. - While brominated lipids constituted only 8% of the initial chemical library, they accounted for more than half of the top-performing candidates identified by the system. - Future plans for the LUMI-lab platform include expanding its optimization capabilities to simultaneously screen for multiple clinically relevant properties beyond delivery potency, such as safety, tolerability, and tissue selectivity.
