A Self-Driving Lab Just Rewrote the Rules of mRNA Delivery

A self-driving AI platform has identified a novel lipid class that outperforms the delivery system at the core of Moderna's Covid-19 vaccine

8 Apr 2026

Researchers at the University of Toronto have published findings showing that an AI-powered self-driving laboratory autonomously discovered a new class of materials capable of improving how mRNA medicines enter human cells, without any human direction on what to find.

The system, called LUMI-lab, screened more than 1,700 lipid nanoparticle formulations across ten experimental cycles. It identified brominated lipid tails as a previously unknown structural feature with significant delivery-enhancing properties. The findings appeared in Cell in February 2026, with support from the Acceleration Consortium and GSK Canada.

Brominated lipids had no prior history in mRNA delivery research.

Despite accounting for just 8% of the chemical library available to the platform, they represented more than half of the top-performing candidates across all experimental rounds. Five of the six best formulations matched or exceeded the performance of SM-102, the ionizable lipid used in Moderna's Covid-19 vaccine. In lung cell models, the lead material achieved 20.3% gene editing efficacy, a new preclinical benchmark for that tissue type.

The commercial significance lies partly in how few options currently exist. Only three lipid nanoparticle formulations have received FDA approval, a constraint that has limited the expansion of mRNA therapeutics into new disease areas and tissue types. LUMI-lab was pretrained on more than 28 million molecular structures, giving it the chemical range to probe territory that conventional screening rarely reaches.

The research team plans to extend the platform's brief, targeting simultaneous optimisation across safety, tolerability, and tissue selectivity, the full profile required before any candidate can enter clinical development.

Whether autonomous discovery at this scale can consistently translate into clinically viable compounds remains an open question. The gap between a strong preclinical result and a regulatory-approved therapeutic is wide, and the brominated lipid class has yet to be tested beyond cell models. As mRNA platforms advance into oncology, metabolic disease, and rare genetic conditions, the pressure to find superior delivery vehicles will intensify. LUMI-lab suggests machines may take the lead in that search. How far they can go before human judgment becomes indispensable again is less certain.