Other work is improving our understanding of how the immune system interacts with other bodily systems in intricate ways. A seemingly simple infection might set off a complex cascade of changes within the body, for example, especially in conditions like long COVID. Conversely, children’s plasma proteome is remodelled by exposure to malaria, leading to natural immunity, and understanding this could lead to new therapeutics.⁵

Bringing such distinct datasets together is challenging but offers the potential for great insights.⁶ It has become possible to predict COVID-19 mortality solely by analysing gene expression in the patient’s blood, for example.⁷ Similarly, rich biological datasets are helping to explain long-established but mysterious patterns in infection biology. For instance, specific changes in the immune system may account for the much higher death rates from COVID-19 among the elderly.⁸

AI is mining insights from the vast datasets generated by research. For example, machine-learning models can predict how virus proteins will interact with host proteins,⁹ and forecast long-COVID outcomes based on immunological data.¹⁰ However, AI expertise and infrastructure is currently centred in the rich West: there is an urgent need to expand AI access in low and middle income countries, where it can enable biomedical progress even in low-resource settings.

Basic knowledge of microbial life also needs to be improved. Some potential zoonotic pathogens, like H5N1 bird flu, are fairly well-characterised.¹² However, the majority of bacteria and viruses have not been studied at all. Researchers are working to change this, for instance by bringing all vertebrate-virus associations together in one database¹³ and by using AI to help document unknown viruses.¹⁴

With this knowledge base in place, it should be possible to predict which pathogens are likely to spill over into the human population and how dangerous they would be if they did.¹⁵ AIs trained on these datasets could help make such predictions, leading to early-warning systems.¹⁶

However, truly reliable predictions require a recognition that outbreaks are not caused just by highly pathogenic organisms but by human disturbance of ecosystems that allows novel diseases to come into contact with our populations. The One Health framework offers a set of tools with which to attempt this.¹⁷ Some models of this type have been developed.¹⁸ However, the One Health approach remains under-used: for example, there is little coordination between surveillance of animal diseases and of human diseases.

Traditional pathogen surveillance methods such as contact tracing can now be combined with other data streams, from genomics²⁰ to social media. For instance, improvements in DNA sequencing mean it is now feasible to rapidly reconstruct how a disease outbreak occurred, including tracing it back to source.²¹ An improved understanding of human factors that affect spread,²² such as malnutrition and genetic vulnerabilities, could significantly aid the tracking of epidemics.²³

However, successfully integrating and using these datasets remains a challenge,²⁴ especially in low- and middle-income countries where resources are limited. It is likely that our initial attempts to predict the course of outbreaks will underperform due to siloed systems, data latency, and interoperability barriers. While such systems may nevertheless offer some insight, it is likely to be too coarse or arrive too late to be used as the basis for effective action.²⁵

Several major opportunities exist. An improved understanding of the interactions between pathogens, which affect when and where outbreaks arise, could enable better predictions. A biobank of infection samples from patients would be a valuable resource for both experimental and computational research. And there is considerable potential to train AI on outbreak datasets and use it to make predictions,²⁶ but so far little has been done — in part due to data-access limitations.

The most effective data ecosystems are likely to be decentralised and trust-based, rather than centralised and coercive. West Africa is a good prototype: pathogen surveillance systems there blend formal and informal health intelligence. Truly resilient systems will arise from diversity, redundancy and local agency, not from top-down architectures. One possible model is to create regional “epidemic foresight nodes” that can detect patterns and simulate response scenarios.

Some new therapeutics have already found widespread use. mRNA vaccines were deployed in vast numbers against covid-19, after decades in development hell. So were monoclonal antibodies. Gene editing for inherited diseases is also increasingly mainstream.

Other promising approaches remain largely untapped. One such example is phage therapy — using viruses called bacteriophages to treat bacterial infections.²⁷ Phage therapy could be used to treat antimicrobial-resistant bacteria, saving many lives.²⁸ A key challenge is to better characterise the interactions between the viruses and their bacterial hosts,²⁹ which will help ensure that the chosen phages actually destroy their target organism.³⁰

A second avenue of attack is to hamper the evolution of antimicrobial resistance, ensuring that antibiotics remain useful for longer. This requires an improved understanding of bacterial evolution.³¹ Key challenges include predicting the behaviour of mobile genetic elements³² and improving our understanding of how drug treatments trigger the evolution of resistance.³³,³⁴

Techniques such as cryogenic electron microscopy and AI are enabling rapid progress in our understanding of molecular interactions such as protein-protein interactions.³⁵ This promises a multitude of new therapeutic targets that could be targeted rapidly using the tools of synthetic biology. Existing drugs could also be repurposed.³⁶

Finally, the rapid emergence of new pathogens is spurring attempts to develop broad-spectrum treatments such as a pan-coronavirus vaccine.³⁷ Multiple methods are being pursued: for instance, it may be possible to develop a “universal antibody vaccine” based on monoclonal antibodies.³⁸