Reviewed by Lexie CornerJan 29 2025
Researchers at the Ragon Institute and MIT's Jameel Clinic have developed MUNIS, a deep-learning model that significantly enhances T cell epitope prediction. This model could potentially accelerate vaccine development for infectious diseases. Their findings were published in Nature Machine Intelligence. This collaboration between MIT’s Jameel Clinic and the Ragon Institute represents a key advancement in using artificial intelligence (AI) for T cell vaccine candidate design.
Image Credit: Ursa Design/Shutterstock.com
MIT Professor Regina Barzilay, Ph.D., AI lead at the Jameel Clinic for AI and Health, and Ragon faculty member Gaurav Gaiha, M.D., D.Phil., introduced MUNIS as a predictive framework capable of identifying CD8+ T cell epitopes with unprecedented accuracy. This breakthrough could streamline vaccine design for various infectious diseases by improving the speed and precision of epitope mapping.
A milestone achievement of the Mark and Lisa Schwartz AI/ML Initiative at the Ragon Institute, this project integrates translational immunology, machine learning, and AI to tackle global infectious diseases. The initiative was made possible through the support of Mark Schwartz, Chair of the Ragon Institute Board, and his wife, Lisa Schwartz.
The research team, led by co-first authors Jeremy Wohlwend, Ph.D., and Anusha Nathan, Ph.D., sought to overcome a longstanding challenge in vaccine development: the efficient and accurate identification of T cell epitopes in pathogenic proteins. Their approach combined the Gaiha Lab’s expertise in T cell immunology with the Barzilay Lab’s advancements in AI-driven modeling. T cell epitopes—discrete regions of an antigen recognized by immune cells—play a fundamental role in eliciting targeted immune responses.
Traditional epitope prediction methods often suffer from limited speed and accuracy. By using state-of-the-art AI architectures and a curated dataset of over 650,000 unique human leukocyte antigen (HLA) ligands, MUNIS significantly outperformed existing models. The system was validated using experimental data from influenza, HIV, and Epstein-Barr virus (EBV), where it successfully identified novel immunogenic epitopes in EBV.
Notably, MUNIS demonstrated accuracy on par with experimental stability assays, a widely used method for epitope validation. This suggests its potential to streamline vaccine development by reducing the reliance on labor-intensive laboratory experiments while maintaining high predictive performance.
This is our first paper at the intersection of AI and immunology. Through this collaboration with Dr. Gaiha and team, we learned a lot about this fascinating field and are excited about the immense possibilities in using AI algorithms to model the intricacies of the immune system.
Regina Barzilay, Ph.D, AI Lead, Jameel Clinic for AI and Health
The development of MUNIS was made possible through close collaboration between immunologists and computer scientists, combining domain expertise in immunobiology with advanced computational modeling. This interdisciplinary approach was essential in ensuring the tool’s ability to navigate the complexities of biological data with high precision.
This is a wonderful application of artificial intelligence that benefited greatly from insights shared by both computer scientists and immunologists. The credit lies with the initiative for bringing us together, which has led to the creation of an exciting new tool for immunology and vaccine design.
Gaurav Gaiha, MD, D.Phil., Faculty Member, Ragon Institute
Beyond vaccine research, MUNIS has broader implications for immunotherapy and autoimmune disease studies. By providing a robust framework for predicting immunodominant epitopes—those most readily recognized by the immune system—it lays the foundation for applications in cancer T cell immunotherapy and autoimmune disease modeling. As the global scientific community continues to address emerging infectious threats, tools like MUNIS offer a scalable solution to enhance preparedness and response strategies.
This advancement underscores the Ragon Institute’s commitment to integrating immunology and technology to drive innovation in global health and disease prevention.
Journal Reference:
Wohlwend, J., Deep learning enhances the prediction of HLA class I-presented CD8+ T cell epitopes in foreign pathogens. Nature Machine Intelligence. doi.org/10.1038/s42256-024-00971-y