Cary Institute hoping machine learning could help identify future disease outbreaks
The Cary Institute of Ecosystem Studies in Dutchess County recently held a virtual talk about using machine intelligence to predict the next pandemic.
Zoonotic diseases like coronaviruses, Ebola, Zika and Monkeypox spread from wild and domestic animals to humans. Researchers say millions of people die each year as a result of such outbreaks, which have been increasing in frequency.
The Cary Institute held a talk featured disease ecologist Dr. Barbara Han in conversation with Cary President Dr. Joshua Ginsberg. Han has led research on global patterns of zoonotic disease in mammals and currently focuses on ecology, computing, and global public health. Han says COVID transformed society.
"Where are we right now? We're on Zoom, we would ordinarily be doing this in person, right? An auditorium full of people unmasked," said Han. "Yeah, disease ecology, I think is now surprisingly, really front and center. It's been part of the national dialogue in a way that I have never seen before in my career."
Han also points out zoonotic diseases are appearing in humans at a more rapid rate than in the past.
"Even after you correct for things like the advent of the Internet, where information about diseases spreads faster and becomes available more much more easily, even after controlling for that, the incidence of these diseases and the emergence events has gone up since the 1980s," Han said. "So I think this is a real phenomenon."
What if we could forecast pathogen spillovers and intervene before they evolve into new variants that harm people, other animals, or ecosystems? That’s what Han is studying, using artificial intelligence machine programs to look for patterns and make predictions.
"Train an algorithm to distinguish between, you know, the characteristics of something that's a disease carrier versus something that's not," Han said. "So that's one really simple that was the very first way that we try to apply machine learning to this question about does infectious disease spillover, and which species should we be concerned about? Are there ways that we can, what can the algorithms suggest about these species that will allow us to then go out and test hypotheses in a smart way.”
Han's computations have identified mammal species most likely to harbor COVID-19 and transmit variants to humans. She is helping advance an infectious disease surveillance system with real-time data streams in an attempt to pinpoint when, where and why zoonotic diseases jump from animals to people. Han notes the computer programs depend on human legwork.
“But we still really depend on those surveillance efforts, like the folks that go out in the field and catch the bats and test the bats," said Han. "That is really incredibly hard work. And that stuff still remains underfunded despite, despite the obvious need for those things. So I think machine learning, it really depends on the data, the quality of the data that you have, and you can't expect to get anywhere with machine learning and AI like, it's a magic wand or something without the hard work of the folks that are collecting and curating and making this data available to the world.”
Han says as people and wildlife interactions increase, so does the likelihood of pandemics and disease spillover.
"The system that we're probably most familiar with is Lyme disease, because the story there is that, by reducing biodiversity, you're actually leaving behind the species that do quite well at amplifying the Borrelia pathogen and also allowing for the ticks that are responsible for feeding on humans most frequently, to persistent in high numbers," Han said. "And so that has been studied for decades, the ecological data are so solid, they're the studies that inspired me to go into the field that I'm in now. But there are other systems in which the answers to that question are quite complicated, right? Because in some cases, you have higher biodiversity. And that means you might have more individuals of a species that is actually super competent for a pathogen. And then you have to think about what are the interactions between the species? And how do those interactions change when a human, when a development moves in and changes the interactions.”
Han says society cannot let its guard down when it comes to transmissible diseases.
"Everyone says ‘well things are going to be different now we've had COVID. It's a learning experience.’ And is it is it going to be a learning experience," asked Han.
“Well, monkeypox wouldn't suggest we learned a lot," replied Ginsberg.
"I know, no," Han said. "And are people going to say we had a pandemic next year? Like, are people gonna say, ‘ah we've been through this, whatever’ now, It’s like, are we going to snap to attention? And we know what to do now like, Hong Kong was so good at this, but SARS two after SARS one, right, they knew that masks worked. And it wasn't a social taboo to be wearing a mask. It was just something that you did. And I that's what I worry about the most, I think. The technology I feel so optimistic about that the collaborations with computer scientists are amazing. And with virologists, I mean, I think all of the motivations are there. It's the complacency in this, in our inability to sort of map this onto social behavior and how to do the communications right. That's what keeps me up at night."
Han believes education and dissemination of information are critical any time people are confronted by a new pathogen.