MIT team uses AI to make mRNA vaccines that survive a year without freezing

Engineers at the Massachusetts Institute of Technology have redesigned the fatty shell that carries mRNA vaccines so that it no longer needs a freezer. Using a machine-learning algorithm built with computer scientists at MIT's Computer Science and Artificial Intelligence Laboratory, the team produced lipid nanoparticle formulations that stay stable at room temperature for up to a year and at nearly body temperature for two months, according to a study published Monday in Nature Biotechnology.

Current mRNA vaccines, including the Moderna and Pfizer-BioNTech COVID-19 shots, rely on lipid nanoparticles, or LNPs, tiny fat-based capsules that shield fragile mRNA strands and ferry them into cells. Those capsules typically break down unless kept at -20 to -80 degrees Celsius, a requirement that has forced clinics worldwide to invest in specialized freezers and cold-chain logistics. The new work, described in a write-up from MIT's news office, set out to remove that constraint entirely.

What the researchers did

The team, based primarily out of MIT's Koch Institute for Integrative Cancer Research, screened nearly 50 FDA-approved excipients — stabilizing ingredients such as sugars, salts and polymers — for their ability to protect lipid nanoparticles from degrading. They measured each candidate's effectiveness using a bioluminescence assay built around firefly luciferase, then narrowed the field to five promising compounds. From there, a machine-learning algorithm developed with the lab of CSAIL's Mina Konaković Luković predicted which ratios of those five excipients would produce the most stable formulation, with each round of lab testing feeding new data back into the model.

• Vaccines remained stable at room temperature for up to one year.
• They held up at roughly 98 degrees Fahrenheit (37 degrees Celsius) for two months.
• Nearly 50 FDA-approved excipients were screened; five were carried forward for optimization.
• Mice given the heat-stable formulation, after storage, mounted immune responses matching a standard Moderna-type vaccine.

The cold-storage problem is not new. During the COVID-19 pandemic, the earliest mRNA vaccines from Pfizer-BioNTech required storage as low as -70 degrees Celsius, colder than a typical medical freezer, and health systems in dozens of countries reported shipments spoiling or clinics turning away doses because they lacked the specialized equipment to keep them viable. Later formulations eased that requirement somewhat, but mRNA vaccines have continued to need standard freezer temperatures that many rural and low-income settings still struggle to guarantee around the clock.

Why it matters

The stakes are largely about reach. A peer-reviewed analysis of the global mRNA supply chain has documented how ultracold storage requirements strain clinics in low- and middle-income countries, where unreliable electricity and scarce ultra-low-temperature freezers can let vaccines drift outside safe temperature ranges before they ever reach a patient's arm. A formulation that survives a year on a shelf, rather than weeks in a freezer, would let health workers skip much of that infrastructure, particularly in rural clinics and during emergency outbreak response where cold-chain logistics are often the limiting factor rather than vaccine supply itself.

The MIT team, whose work was partly funded by the Bill & Melinda Gates Foundation, also tested whether the new stability would hold up outside a syringe. Using the same excipient formulation, they built solid microneedle patches loaded with a SARS-CoV-2 antigen; in animal testing, the patches produced immune protection comparable to an injected vaccine. The group additionally showed the algorithm could stabilize a separate, Pfizer-style lipid nanoparticle recipe using the same class of excipients, suggesting the approach is not limited to one company's formulation.

What the researchers are saying

Ana Jaklenec, a principal investigator at the Koch Institute and one of the paper's senior authors alongside David H. Koch Institute Professor Robert Langer, said the appeal of the algorithm was how little data it needed to work.

"The real beauty of this algorithm is that we can use it with small data sets. It's really hard to run thousands of experiments, so this algorithm allows us to more easily achieve formulations with features that we want — in this case, stability," Jaklenec said.

Khanh Tran, a postdoctoral researcher and one of the paper's lead authors along with graduate student Jinbi Tian, said the AI-guided process compressed work that had previously dragged on for months. Earlier attempts to find a stable formulation by testing combinations one at a time took "several months" without reaching full stability, he said, while the machine-learning approach reached a workable formulation in a matter of weeks. Konaković Luković, an assistant professor of electrical engineering and computer science, said the model converged on stable formulations "in just a handful of iterations, rather than the exhaustive search that would normally be required."

What happens next

The findings are still at the animal-testing stage, and researchers caution that the usual gap between promising mouse data and an approved product remains: human trials, manufacturing scale-up and regulatory review would all be required before any heat-stable version of an existing vaccine reaches clinics. The MIT group says it plans to keep developing the microneedle-patch delivery format and to test whether the same excipient-optimization approach can stabilize other RNA-based therapies beyond vaccines, including experimental treatments for cancer that rely on the same lipid nanoparticle delivery system.

The researchers also say the microneedle-patch results point toward a longer-term shift in how vaccines might be administered altogether. A patch that a patient or minimally trained worker can apply without a syringe, refrigeration or medical waste disposal would remove several logistical steps at once, from cold storage to trained staff to needle disposal. For now, that remains a research prototype rather than a product, but the team frames the new formulation work as a foundation other labs and vaccine developers could build on rather than a single fix tied to one company's vaccine.