Accelerated discovery of thermostable vaccines using data-efficient AI
- 1MIT
Abstract
The inherent instability of mRNA- lipid nanoparticles (LNPs) necessitates ultra-cold storage, creating significant barriers for global distribution and limiting their broader application in advanced delivery systems. Solid-state, water-free formulations offer a promising solution by enhancing thermostability and enabling integration into emerging delivery modalities such as microneedle (MN) patches. Prior efforts to stabilize mRNA-LNPs have been constrained by narrow formulation scope and low-throughput screening methods. Here, we introduce AGENT (Algorithm-Guided Experimental design for lipid Nanoparticle Thermostabilization), an AI-driven framework that couples high-throughput experimentation with Bayesian optimization to rapidly identify thermostable mRNA-LNP formulations. Manual exploration of the formulation space required months of screening and yielded suboptimal candidates. In contrast, AGENT extracted maximal information from sparse experimental datasets, enabling efficient formulation optimization in only six iterations completed within one month. Using AGENT, we stabilized mRNA vaccines with diverse LNPs, including those in clinical use, into solid state formulations that retained 100% bioactivity after storage at 37 degree C for over two months. The thermostable vaccines induced antigen-specific IgG and germinal center B cell responses that were non-inferior to those elicited by freshly prepared soluble vaccines. The solid-state formulations were further incorporated into dissolvable MN patches and administered to rodents and nonhuman primates, yielding comparable neutralizing antibody titers compared to conventional intramuscular delivery of fresh vaccines. To our knowledge, this study presents the first demonstration of AI-driven design of thermostable RNA vaccines, offering a scalable, cold-chain-free solution for global immunization. By addressing both stability and delivery challenges, AGENT provides a potentially transformative platform for developing accessible next-generation therapeutics.
Related
- Expanding the Thermomyces lanuginosus lipase repertoire: genome-guided discovery and characterisation of three thermostable lipolytic enzymes
- Efficient and precise discovery of LTR-retrotransposon families using HiRepEat
- Directed evolution of a thermostable family-A DNA polymerase into a reverse transcriptase for 2'-modified RNA
- How thermostable direct haemolysin (TDH) diverges from TDH-related haemolysin (TRH)? Reassessing haemolysins in Vibrio parahaemolyticus through functional and structural representation learning
- Machine-learning-guided enzyme discovery and redox-system engineering for efficient production of the nylon monomer methyl 12-aminododecanoate in Escherichia coli
BibTeX
@misc{tian2026accelerated,
title = {{Accelerated discovery of thermostable vaccines using data-efficient AI}},
author = {J. Tian and K. T. M. Tran and B. H. Pogostin and O. Sheridan and S. Mursalova and A. H. Lee and S. Liu and J. Hamkins and D. Antov and A. L. Power and Z. S. Dash and D. Yun and M. Konakovic Lukovic and R. Langer and A. Jaklenec},
howpublished = {bioRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.17.752370},
url = {https://www.biorxiv.org/content/10.64898/2026.09.17.752370v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Accelerated discovery of thermostable vaccines using data-efficient AI
AU - J. Tian
AU - K. T. M. Tran
AU - B. H. Pogostin
AU - O. Sheridan
AU - S. Mursalova
AU - A. H. Lee
AU - S. Liu
AU - J. Hamkins
AU - D. Antov
AU - A. L. Power
AU - Z. S. Dash
AU - D. Yun
AU - M. Konakovic Lukovic
AU - R. Langer
AU - A. Jaklenec
PY - 2026
DA - 2026/09/17
PB - bioRxiv
DO - 10.64898/2026.09.17.752370
UR - https://www.biorxiv.org/content/10.64898/2026.09.17.752370v1
AB - The inherent instability of mRNA- lipid nanoparticles (LNPs) necessitates ultra-cold storage, creating significant barriers for global distribution and limiting their broader application in advanced delivery systems. Solid-state, water-free formulations offer a promising solution by enhancing thermostability and enabling integration into emerging delivery modalities such as microneedle (MN) patches. Prior efforts to stabilize mRNA-LNPs have been constrained by narrow formulation scope and low-throughput screening methods. Here, we introduce AGENT (Algorithm-Guided Experimental design for lipid Nanoparticle Thermostabilization), an AI-driven framework that couples high-throughput experimentation with Bayesian optimization to rapidly identify thermostable mRNA-LNP formulations. Manual exploration of the formulation space required months of screening and yielded suboptimal candidates. In contrast, AGENT extracted maximal information from sparse experimental datasets, enabling efficient formulation optimization in only six iterations completed within one month. Using AGENT, we stabilized mRNA vaccines with diverse LNPs, including those in clinical use, into solid state formulations that retained 100% bioactivity after storage at 37 degree C for over two months. The thermostable vaccines induced antigen-specific IgG and germinal center B cell responses that were non-inferior to those elicited by freshly prepared soluble vaccines. The solid-state formulations were further incorporated into dissolvable MN patches and administered to rodents and nonhuman primates, yielding comparable neutralizing antibody titers compared to conventional intramuscular delivery of fresh vaccines. To our knowledge, this study presents the first demonstration of AI-driven design of thermostable RNA vaccines, offering a scalable, cold-chain-free solution for global immunization. By addressing both stability and delivery challenges, AGENT provides a potentially transformative platform for developing accessible next-generation therapeutics.
ER -