MIT team uses AI to make mRNA vaccines that stay stable for a year at room temperature

Vaccine particles that normally have to be frozen stayed stable for up to a year at room temperature in a new recipe worked out with the help of an AI algorithm, according to MIT News. The same particles held up for two months at nearly 100 °F (about 37 °C). The study, led by Ana Jaklenec and Robert Langer of MIT's Koch Institute for Integrative Cancer Research, was published in Nature Biotechnology at the end of September.

All results so far come from cell experiments and mice. No person has received the new formulation.

Ultra-cold storage kept the first mRNA vaccines out of many countries

mRNA vaccines carry their genetic instructions inside tiny fat bubbles called lipid nanoparticles. Those particles break down quickly when they warm up, so today's RNA vaccines are stored at -20 to -80 °C. That requirement is expensive and hard to meet in places without reliable freezers or electricity.

The problem became visible during the Covid-19 rollout. When the Pfizer-BioNTech vaccine was first listed for emergency use by the World Health Organization, its need for ultra-cold shipping limited its distribution through the global COVAX scheme to just 18 countries by mid-2021, the vaccine alliance Gavi reported. Very few countries could handle that kind of cold chain, Gavi wrote.

An algorithm that works with a handful of experiments

The MIT team focused on excipients, the inactive ingredients added to the particles. They first screened nearly 50 excipients already approved by the U.S. Food and Drug Administration, using mRNA that makes cells glow like a firefly. The more light the cells gave off, the better the mRNA had been protected.

Five excipients made the cut. The researchers then let an algorithm, developed with Mina Konaković Luković's group at MIT's Computer Science and Artificial Intelligence Laboratory, suggest ratios for particles similar to Moderna's Covid-19 vaccine. They tested two recipes at a time, fed the results back and repeated the cycle. The whole process took a few weeks, and the finished particles were vacuum-dried before storage.

“The real beauty of this algorithm is that we can use it with small data sets.” — Ana Jaklenec, principal investigator, MIT Koch Institute for Integrative Cancer Research

Doing it by hand had not worked. "Before we implemented the AI algorithm, we spent several months testing different combinations and also doing the prescreening of all the excipients that we could find, but nothing would get us to 100 percent stability," said Khanh Tran, a postdoc and one of the two lead authors.

Mice responded as strongly as to a Moderna-like vaccine

In mice, Covid-19 vaccines packed into the new particles produced immune responses equivalent to those from particles similar to the original Moderna formulation, even after long-term storage. Microneedle skin patches made with the recipe and carrying a SARS-CoV-2 antigen triggered a response similar to injected RNA vaccines. The same five ingredients, mixed in a different ratio, also stabilized a particle similar to Pfizer's Covid-19 vaccine.

Earlier heat-stable particles from the same lab used polymers, which made them different from the vaccines regulators had already approved. Sticking to approved ingredients this time was a deliberate choice, according to MIT.

Lead author Jinbi Tian, a graduate student, said the approach could reach beyond vaccines to "therapeutics or advanced drug-delivery platforms like controlled-release particles or microneedle patches." The work was partly funded by the Gates Foundation. The paper does not give a timeline for testing the formulation in people.