Pandemic scientists combine AI, ecology to predict future

On a balmy Thursday evening in November, staying indoors to attend the Cary Institute of Ecosystem Studies’ virtual talk, “Predicting the Next Pandemic with AI and Ecology” featuring Barbara Han, a disease ecologist, proved interesting and informative — even as the science discussed spelled out a difficult road ahead.

The conversation between Han, of the Cary Institute, and Joshua Ginsberg, its president, covered the broad range of Han’s groundbreaking work at the intersections of machine learning, disease ecology and social epidemiology.

Her research focuses on deploying artificial intelligence to study “zoonotic” diseases, which are diseases that spill over from the animal kingdom into human populations.

The end result is that the AI is able to accurately predict what species might be “strong carriers” of different kinds of zoonotic diseases.

Approximately two-thirds of emerging diseases in humans are zoonotic, meaning they arise in human populations as a result of the spillover effect. Notable examples include Lyme disease, Ebola virus and avian flu. Understanding how this phenomenon works, and what species in the animal kingdom are strong carriers, has meaningful implications for our ability to predict and combat future pandemics.

In the case of COVID-19, however, Han was the bearer of some bad news. One of the predictions made in Han’s research, published in November 2021 by The Royal Society, is that white-tailed deer would be strong carriers of COVID-19. This prediction was confirmed by independent field studies, which collected hard evidence that the virus is  freely circulating in those populations

This fact has profoundly negative implications for our ability to manage COVID outbreaks, Han suggested. Even if we were somehow able to reduce or eliminate COVID-19 in human populations, there exists a vast and stable reservoir of the virus in white-tailed deer — an overabundant species with which humans have high contact. This suggests a high likelihood that COVID could easily spill back over into human populations. Ginsberg spoke to first encountering this information in an early draft of Han’s paper, and he recalled the sinking feeling it gave him.

“I remember reading it and being more depressed about the pandemic — because it’s really clear, it’s not going away.… As a zoonotic disease, it is now in other species, and it’s evolving in those other species, and inevitably it will eventually spill back from one of them to us.”

The crux of Han’s research methods lies in training an AI to identify what traits make an animal species a strong carrier of diseases. For example — rodents, such as rats, are prolific in this regard. What traits contribute to that? High birth rates? Fast metabolisms?

Once a program knows these things, it can begin to apply that knowledge to much larger datasets than are typically approachable by human researchers. Through this process, an AI can make hard-to-find connections, and out of them create highly accurate predictions of what animals, from a wide pool, might be strong carriers of specific diseases and of categories of diseases.

Armed with a deeper understanding of zoonotic diseases, and with access to the powerful analytic capabilities of AI, Han expressed optimism in our ability to marshal available data and use it to help manage future pandemics. However, when asked what keeps her up at night, Han was honest about where her concerns reside.

“What keeps me up at night is our complacency …Say we had a pandemic next year, are people going to say, ‘Ah, we’ve been through this, whatever. Masks, I’m not gonna.’ Are we going snap to attention, and [say]: We know what to do now? That’s what I worry about the most, I think.”

The Predicting the Next Pandemic talk is available at www.caryinstitute.org/events.