Researchers from Oxford and the Pandemic Sciences Institute collaborate with colleagues from around the world in the first study of its kind to be published in Nature, outlining how AI can revolutionize the study of infectious diseases and save more lives.
Image Credit: Pandemic Sciences Institute
For the first time, the study describes how developments in AI can hasten discoveries in the study of infectious diseases and the response to outbreaks.
The study focuses on safety, accountability, and ethics in the deployment and use of AI in infectious disease research. It was released after the AI Action Summit and during the growing worldwide debate on AI investment and regulation.
The study, a collaboration between PSI and the University of Oxford scientists and colleagues from academia, industry, and policy organizations across Africa, America, Asia, Australia, and Europe, calls for a transparent and cooperative environment regarding datasets and AI models.
AI in medicine has so far mostly been used to improve individual patient care, such as by supporting clinical treatment decisions, improving clinical diagnostics, or practicing precision medicine.
Instead, the application of AI to population health is examined in this review. Even with insufficient data, there is a significant bottleneck until recent developments in AI approaches are performing better and better, according to the study. AI tools are finding new applications in both high- and low-income nations to enhance health because of their improved performance on noisy and sparse data.
In the next five years, AI has the potential to transform pandemic preparedness. It will help us better anticipate where outbreaks will start and predict their trajectory, using terabytes of routinely collected climatic and socio-economic data. It might also help predict the impact of disease outbreaks on individual patients by studying the interactions between the immune system and emerging pathogens.
Moritz Kraemer, Study Lead author and Professor, Pandemic Sciences Institute, University of Oxford
Kraemer added, “Taken together and if integrated into countries’ pandemic response systems, these advances will have the potential to save lives and ensure the world is better prepared for future pandemic threats.”
The Opportunities Ahead
The study found the following opportunities for AI and pandemic preparedness:
- Promising developments in enhancing current disease spread models to make them more realistic, precise, and robust
- Finding regions with high transmission potential has advanced so that scarce healthcare resources can be distributed as effectively as feasible
- Possibility of enhancing genetic information in disease surveillance, which would speed up the creation of vaccines and the discovery of novel variations
- Potential can assist in predicting the characteristics of novel infections, figuring out their attributes, and determining the likelihood of cross-species jumps
- Identifying potential new pathogen variants, such as influenza and SARS-CoV-2, and determining which vaccines and treatments will have the greatest effect on lessening their effects
- Potential AI-assisted integration of data from individual-level sources, such as wearable devices like heart rate and step counts, with data from the population to better identify and track outbreaks
- AI can improve capacity in settings that most need these tools by establishing a new interface between healthcare professionals with little training and highly technical science
However, the impact of AI advancements will not be uniform across all aspects of pandemic preparedness and response. For instance, advances in foundational models may only offer slight improvements over current methods for modeling the rate of pathogen spread, while protein language models have enormous potential for accelerating the understanding of how virus mutations can affect disease severity and spread.
Human Feedback and Global Collaboration
While integrating human feedback into AI modeling workflows may help overcome current limitations, scientists caution against assuming that AI alone will solve infectious disease challenges.
The authors' main concerns are the quality and representativeness of training data, the restricted availability of AI models to the general public, and possible hazards related to the use of black-box models for decision-making.
While AI has remarkable transformative potential for pandemic mitigation, it is dependent upon extensive worldwide collaboration and from comprehensive, continuous surveillance data inputs.
Eric Topol, MD, Study Author and Professor, University of Oxford
Eric Topol is also the Founder And Director of the Scripps Research Translational Institute.
Infectious disease outbreaks remain a constant threat, but AI offers policymakers a powerful new set of tools to guide informed decisions on when and how to intervene.
Samir Bhatt, Study Lead Author, University of Copenhagen
The authors advocate for strong partnerships between academia, industry, society, and government to develop sustainable and useful models for improving human health. They also propose strict criteria for evaluating AI models.
Journal Reference:
Kraemer, M. U. G., et al. (2025) Artificial intelligence for modelling infectious disease epidemics. Nature. doi.org/10.1038/s41586-024-08564-w.