1 October 2026

 

More than two months at 37°C is a demanding storage test for an mRNA vaccine. Experimental formulations developed by researchers at the Massachusetts Institute of Technology survived that exposure with their measured biological activity intact.

Published in Nature Biotechnology on 28 September, the study uses artificial intelligence to help identify recipes that protect mRNA and its delivery particles when dried into a solid form. The researchers report retaining 100% bioactivity under the tested conditions. That figure describes performance in their biological tests, rather than effectiveness at preventing disease. 

The practical attraction is clear. Vaccines must remain usable throughout the journey from factory to patient. Improving their tolerance of heat could give distribution teams more flexibility when refrigeration is difficult, transport is delayed or vaccination takes place outside a clinic.

The research connects three stages of development: finding a protective formulation, measuring its performance after storage and testing whether it can stimulate an immune response. Turning those results into a deployable vaccine will require manufacturing and clinical work.

 

What needs stabilising in an mRNA vaccine?

 

Messenger RNA, or mRNA, provides instructions that cells use to make a protein. In a vaccine, that protein gives the immune system a target to recognise. Lipid nanoparticles, microscopic carriers made from fatty molecules, protect the instructions and help deliver them into cells.

Both components must remain functional. Preserving the RNA is insufficient if its carrier can no longer deliver the cargo effectively.

The team, including joint first authors Jinbi Tian and Khanh T. M. Tran and supervisors Robert Langer and Ana Jaklenec, investigated formulations using two clinically relevant lipid nanoparticle compositions. These were representative of the SM-102-based and ALC-0315-based systems used in Moderna and Pfizer–BioNTech vaccines respectively. The experiments do not establish that those companies’ existing commercial products can tolerate the same storage conditions. 

Removing water can improve stability, but drying introduces its own stresses. A protective formulation has to help the vaccine material survive processing as well as subsequent storage.

Finding that recipe is a complex search. Ingredients and their proportions interact, so testing an additive on its own reveals only part of the picture. A combination that performs well at one concentration may behave differently when another component changes.

 

How AGENT guides vaccine formulation

 

The researchers developed AGENT, short for Algorithm-Guided Experimental design for lipid Nanoparticle Thermostabilization. It combines high-throughput experimentation with Bayesian optimisation, a statistical method for guiding searches where each test consumes time and resources.

Bayesian optimisation builds a predictive model from completed experiments. It estimates how untested combinations might perform and how uncertain those predictions are. Those estimates help select the next experiments, balancing promising candidates with combinations that could reveal useful new information. 

Consider a laboratory with several formulation ingredients to adjust. Working systematically through every combination could quickly become impractical. An optimisation framework uses the results already obtained to narrow the next selection, updating its recommendations as evidence accumulates.

AGENT applies that logic to vaccine formulation. Researchers prepare and test material, feed measurements into the framework and use its recommendations to inform another round. The algorithm guides the search; laboratory experiments establish whether the selected recipes work.

The study reports six optimisation rounds completed within one month. This is the duration of the formulation optimisation, rather than the entire research programme or a timetable for bringing a vaccine to market. 

The value of this approach lies in learning efficiently from limited experimental data. Its recommendations remain bounded by the ingredients, conditions and measurements included in the search.

 

What the storage and animal tests showed

 

Storage performance was followed by animal testing. The researchers report antigen-specific immune responses in rodents and non-human primates that were non-inferior to those elicited by intramuscular delivery of freshly prepared liquid vaccines. These responses support further investigation, but do not establish clinical protection in people. 

The work also examined microneedle patches, which introduce vaccine material through the skin using arrays of tiny projections. The accompanying research briefing reports delivery with the thermostable formulations in rodents and non-human primates.

That adds a delivery prospect to the storage result. A compact patch containing heat-tolerant vaccine material could eventually offer vaccination programmes a different format to transport and administer.

It builds on earlier work by Jaklenec and colleagues. In 2023, MIT described a prototype printer producing vaccine-containing microneedle patches that remained active after room-temperature storage and generated immune responses in mice. The new study brings an AI-guided formulation search to this broader effort to make mRNA vaccination more portable.

 

From heat-stable formulation to deployable vaccine

 

A constant-temperature experiment answers a defined question. Distribution introduces changing temperatures, handling, transport delays and exposure to different environments. Developers would need evidence covering the conditions permitted for an eventual product.

The World Health Organization requires vaccine stability to be established through testing. Storage instructions and shelf life must preserve acceptable potency, identity and purity. A successful heat-storage result is therefore one part of a larger evidence package. 

Manufacturing would bring further questions: can the formulation perform consistently across large batches, which drying process is suitable, and what packaging is needed to maintain stability? A patch would also require dependable manufacture and dose delivery.

The economics remain to be demonstrated. Reduced refrigeration demands could create savings, while drying and specialised packaging could add costs. The balance would depend on the product and the distribution system in which it is used.

Human studies would then need to establish safety and clinical performance. The published findings provide a basis for that development, rather than a finished vaccine ready for deployment.

AI’s contribution here concerns an often overlooked part of medicine: preserving a product after it has been made. A more efficient search has produced formulations worth taking further. If they progress, the benefit could be measured in the extra time vaccination teams have to manage a delayed shipment, organise an outreach session or reach patients beyond reliable refrigeration.