A research team at the Massachusetts Institute of Technology (MIT) has developed a novel nanoparticle design system using artificial intelligence, significantly improving the delivery efficiency of RNA vaccines and therapies. The research results were published in the journal Nature Nanotechnology, providing an innovative tool for RNA drug development.

The team created a database containing 3,000 lipid nanoparticle (LNP) formulations and trained a machine learning model named COMET to analyze it. Project leader Professor Giovanni Traverso stated: "This tool can quickly identify the optimal combination of ingredients, greatly shortening the development cycle." Experiments confirmed that the new LNP formulations predicted by the model outperformed existing commercial products in mRNA delivery efficiency.

The system is based on a Transformer architecture and can understand the synergistic effects of different chemical components. Researcher Alvin Chan noted: "Traditional methods can only optimize one compound at a time, whereas COMET can handle complex combinations of multiple components." The model can also predict the best carriers for specific cell types and evaluate the lyophilization stability of nanoparticles.

The technology is currently being applied to research on treatments for diabetes and obesity, including the development of delivery systems for drugs such as GLP-1 analogs. The team plans to further expand the database scale and explore more novel material combinations.