Share Tweet

At VEME 2026 in Stellenbosch, Professors Moritz Kraemer and Houriiyah Tegally explored a central challenge for future outbreak preparedness: how to connect increasingly diverse sources of data to detect emerging threats earlier, understand how they develop, and target surveillance more effectively.

 

When an outbreak first emerges, scientists often have only fragments of information. The pathogen itself may not yet be identified, its growth rate may be unknown, and researchers may have little idea of where it has already spread. 

The keynote Open Lecture, held at Reitz Hall at Stellenbosch University on Wednesday, 9 September, formed part of the 29th International Virus Evolution and Molecular Epidemiology (VEME) Workshop. 

For Professor Moritz Kraemer, Professor of Epidemiology and Data Science at the University of Oxford, one of the central challenges is reducing the time between a pathogen becoming established in a human population and its detection. 

“Only if we move that detection time earlier do we have a real chance to actually contain outbreaks fast and not have to put very stringent interventions and a lot of money on the line to do that,” he said. 

The question, then, is how surveillance systems can be designed to find emerging threats faster. 

Prof Kraemer described work on adaptive surveillance, where information accumulating during an outbreak can help determine where limited testing resources should be deployed next. Rather than relying only on passive systems to detect outbreaks, the aim is to identify an optimal strategy for building a picture of an outbreak as quickly as possible, using as few tests as possible. 

Some of the methods being investigated come from AI research. Active learning approaches developed for applications such as social media algorithms can, Prof Kraemer explained, be reconsidered in a very different context: instead of determining what a person should see in a feed, they could help researchers decide what they should be looking for, and where. 

His team has explored how such approaches could incorporate patterns of human mobility and different layers of surveillance within a country to shorten the time needed to detect a new variant, lineage, or pathogen. 

For Associate Professor in Bioinformatics Houriiyah Tegally, Head of Data Science at Stellenbosch University’s Centre for Epidemic Response and Innovation, the challenge extends across the full trajectory of an outbreak – from the conditions that enable a pathogen to emerge to those that allow it to establish and spread. 

Prof Tegally explained that her team is trying to address an important knowledge gap: how to connect the different pathways that allow pathogens to emerge and develop into outbreaks in urban settings, or even global pandemics. 

While factors such as climate, viral genetics, ecology, and transmission can be studied individually, the challenge is to better understand how they interact. 

Work on Ebola spillover in the eastern Democratic Republic of the Congo illustrates this approach. Existing ecological models had classified the area where an outbreak occurred as being at moderate risk. Prof Tegally and colleagues therefore investigated why spillover had occurred in that particular place. 

Updating the ecological niche model increased the estimated suitability of the area, but the researchers went further, combining ecological information with data on human settlement, forest loss, bushmeat activity, proximity to mining, and conflict. Incorporating these factors improved the ability of the models to identify potential spillover locations. 

The same principle – bringing different forms of information together – runs through work on other emerging pathogens. Prof Tegally described studies combining ecological suitability with population and human movement data, as well as research linking pathogen dispersal with environmental conditions to better understand how outbreaks expand. 

AI offers additional tools for exploring these relationships. Prof Tegally presented work using deep learning to model dengue distribution across different spatial scales, as well as research using satellite imagery to estimate displacement following climate disasters. In the latter study, adding satellite features improved the models, particularly when estimating large-scale displacement. 

But the potential of AI also comes with important limitations. 

“The risk with AI is that we cannot explain what the model tells us,” Prof Tegally said. “One major emphasis when we are using AI in this kind of work is to be able to explain the models and defend the models.” 

That becomes particularly important when a model identifies a particular place as being at high risk. Researchers need to understand why a model reaches its conclusions, particularly when those results could influence surveillance or other decisions. 

Greater use of AI also does not mean abandoning established approaches. Prof Tegally noted that much of her team’s work continues to use traditional mathematical and statistical models, with the choice of method depending on the question, the structure of the data, and how far researchers need to be able to explain the result. 

Prof Kraemer similarly pointed to areas where considerable work remains before AI can deliver more transformative advances in infectious disease research. One challenge is moving beyond individual indicators of emergence towards understanding the full pathway through which a pathogen can move from animal populations to becoming a risk at population level. 

Together, the two perspectives point towards a more connected approach to outbreak science: one that not only gathers more information, but brings different sources of data together to identify where threats may emerge, understand how they are spreading, and determine where surveillance should focus next. 

For Prof Tegally, Africa is particularly well placed to help drive that work. 

“I think that in Africa, we are at a perfect intersection of, sadly, risk, but also of innovation and capacity building to position the continent as a leader in preparedness and in understanding these various different dynamics,” she said. 

The aim, she explained, is ultimately to “close the loop” – linking genomic surveillance and transmission dynamics with risk mapping, and feeding that knowledge back into better surveillance. 

Text: Katrine Anker-Nilssen, Photos: CERI Media, Charlie Sperring

News date: 2026-09-11

Links: