The integration of AI and advanced robotics to create self-driving laboratories (SDLs) is a promising approach for molecular research. A new SDL system called LUMI-Lab combines extensive molecular pre-training, active learning, and robotics, and has discovered that brominated lipids, previously unassociated with mRNA delivery, improve the efficiency of delivering mRNA into human cells.
The study, led by researchers at the University of Toronto’s Leslie Dan Faculty of Pharmacy, was published today in Cell.
The LUMI Lab (Large-scale Unsupervised Modeling followed by Iterative experiments), funded by an AC Translation Research Grant from the University of Toronto’s Acceleration Consortium, integrates a foundational molecular model with automated robotic systems. Unexpectedly to the research team, it discovered a new class of mRNA-enhancing lipids, brominated lipid chains, as a key enhancer for improving transfection efficiency.
“Over ten active learning cycles, the LUMI Lab synthesized and tested over 1,700 novel lipid nanoparticles, discovering brominated, ionizable lipids that deliver mRNA into human lung cells more efficiently than approved comparators,” said Bowen Li, GSK Chair in Pharmaceutics and Drug Delivery at the University of Toronto’s Leslie Dan Faculty of Pharmacy and an associate scientist at the University Health Network’s Princess Margaret Cancer Centre. “The critical advance of this AI-powered system is that it independently identified bromination itself as an important and informative design feature, without prior hypotheses or direction from researchers.”
While mRNA therapeutics are among the fastest-growing classes of drugs, they currently rely on lipid nanoparticles (LNPs) for safe delivery to target areas of the human body. To date, however, only three LNPs have received FDA approval. Platforms like LUMI Lab are expanding the design space by accelerating the discovery of next-generation LNPs needed for new therapeutic applications.
Furthermore, SDLs for drug discovery require large, high-quality datasets to achieve optimal results. In emerging areas such as the development and delivery of mRNA therapeutics, the lack of historical data continues to be a major obstacle. To address this data scarcity in this research field, the team chose a foundational model and pre-trained LUMI on over 28 million molecular structures. This allowed the model to learn general chemical patterns and structures before dedicating itself to more specific tasks.
“When the model is integrated into an active learning framework, it can be continuously optimized in a closed-loop system, further improving its predictive accuracy,” said Li, who also holds the Canada Research Chair in RNA vaccines and therapeutics.
Tested in preclinical models, some of the newly discovered lipids outperformed the lipid used in Moderna’s COVID-19 mRNA vaccine. Although brominated lipids accounted for only 8 percent of the LUMI-Lab’s substance library, they represented over half of the most promising candidates. Brominated lipids also demonstrated a similar safety profile to clinically established lipids, underscoring their potential for future therapeutic development.
“Next, we are expanding LUMI-Lab to optimize multiple clinically relevant properties simultaneously — not just drug release, but also safety, tolerability, and tissue selectivity,” said Li. “By linking AI predictions and automated experiments, we aim to shorten the development cycle of new lipid materials and unlock a significantly larger, evidence-based chemical space for mRNA therapeutics.
