For most of human history, pandemics have arrived as surprises — animals harboring pathogens that suddenly, unpredictably, cross into human lives. A consortium of European researchers, publishing in The Lancet Infectious Diseases, now argues that artificial intelligence paired with the One Health framework — which treats human, animal, and environmental health as inseparable — could compress the window between a spillover's first spark and our awareness of it. Their case is not that technology will end pandemics, but that it might give us something we have rarely had: enough time to act.
AI and One Health approach could strengthen early detection of emerging diseases
Most sparks extinguish, but some ignite fires that spread uncontrollably.
Mark
So the core idea here is that AI can help us spot dangerous viruses in animals before they jump to humans?
Mimi
That's part of it, yes. But it's bigger than just spotting viruses. It's about using AI to analyze massive amounts of data—climate, farming practices, animal movement, human migration—to predict where spillovers are most likely to happen. Then you can focus your surveillance efforts there.
Luke
But how accurate are those predictions? The source talks about AI revealing patterns, but it doesn't say how often those patterns actually point to real threats versus false alarms.
Mimi
That's a fair question. The researchers are working through the VEO consortium to develop these tools, but the comment in The Lancet is more of a vision statement than a report of proven results. They're arguing the approach could work, not that it's already working at scale.
Mark
What about the metagenomic sequencing part? That sounds like it could generate a lot of noise—millions of genetic fragments, most of them harmless.
Mimi
Right. That's where AI helps filter the signal from the noise. When you sequence a sample from wastewater or air, you get millions of fragments. Most match known organisms. But thousands don't. AI can look for patterns in those unknowns and flag which ones might be dangerous.
Luke
Again, though—how? The source says AI can "detect patterns and point to what might be dangerous," but it doesn't explain the mechanism. What patterns? What makes something look dangerous at the genetic level?
Mimi
The researchers mention that viral properties related to human infectivity and transmissibility are embedded in the genetic code. AI protein models can predict how mutations change viral structure, which can affect spread and severity.
Mark
So AI is essentially learning to read the genetic code for danger?
Mimi
In a sense, yes. Though Koopmans is careful to say this is still challenging and mostly potential right now. She talks about "great potential" but also acknowledges we're still in early stages.
Luke
That's important. The comment describes prototypes and possibilities, not deployed systems. And there's the human question too—Koopmans says researchers will need to learn new roles as supervisors and teachers, and they need to figure out how to validate AI output and catch mistakes. That's not a small problem.
Mark
So this is really about building a new kind of partnership between humans and machines?
Mimi
Exactly. The researchers see AI as a "co-scientist" that works alongside human experts, offering analyses they can evaluate. But that requires trust, transparency, and new frameworks for validation. It's not just a technology problem; it's a cultural and institutional one.
El Pulso
- Every pandemic begins as a small, invisible event — a pathogen crossing from animal to human — and by the time we see it, containment is already a race we are losing.
- Researchers from five European institutions warn that climate change, industrial farming, and habitat encroachment are accelerating the frequency of these species-boundary crossings, raising the stakes for early detection.
- AI is being deployed to scan climate records, land use data, transportation networks, and population movements simultaneously, surfacing hidden risk patterns that no human analyst could assemble alone.
- Metagenomic sequencing, guided by AI, can sweep wastewater, air, and food samples for unknown pathogens — and AI protein models can then predict whether a novel virus has the mutations needed to spread dangerously among people.
- The researchers caution that as AI systems grow more capable of running entire research cycles autonomously, scientists must develop new skills as supervisors — able to catch the errors that increasingly sophisticated models may quietly make.
For most of human history, pandemics have arrived as surprises — animals harboring pathogens that suddenly, unpredictably, cross into human lives. A consortium of European researchers, publishing in The Lancet Infectious Diseases, now argues that artificial intelligence paired with the One Health framework — which treats human, animal, and environmental health as inseparable — could compress the window between a spillover's first spark and our awareness of it. Their case is not that technology will end pandemics, but that it might give us something we have rarely had: enough time to act.
A team of researchers spanning Denmark, the Netherlands, Hungary, Italy, and the United Kingdom has published an argument in The Lancet Infectious Diseases: artificial intelligence, combined with the One Health approach to surveillance, could help detect dangerous pathogens before they become pandemics. The core insight is that most pandemic threats begin in animals, and that by the time a spillover becomes visible in human populations, containment is already desperately difficult — a lesson COVID-19 taught at enormous cost.
The authors, including Frank Møller Aarestrup of Denmark's DTU National Food Institute and Marion Koopmans of the Erasmus Medical Centre, emerged from years of collaboration through the VEO consortium, a European initiative building data-driven tools for tracking emerging infectious diseases. They describe spillovers as sparks — most die out harmlessly, but some ignite. The challenge is identifying the dangerous ones early enough to matter.
AI enters this picture by doing what human analysts cannot: simultaneously processing climate records, land use patterns, animal production systems, transportation networks, and population data to map where disease risk is highest. Once those hotspots are identified, metagenomic sequencing can sweep environmental samples — wastewater, air, food — cataloging genetic material and flagging unknown sequences that might signal a novel threat. AI-based protein models can then predict how viral mutations might affect transmissibility and severity, compressing analytical work that once took years into something far faster.
Yet the researchers are careful about what they are and are not claiming. Koopmans acknowledges that as AI systems grow capable of running entire research cycles — generating hypotheses, reviewing literature, analyzing data — scientists will need to become skilled supervisors, able to recognize when these systems err. Aarestrup envisions AI as a recognized collaborator at the research table, not an oracle. The authors conclude that AI's promise for pandemic preparedness is real, but only if it remains a complement to classical surveillance methods rather than a replacement for the human judgment that must ultimately govern it.
A group of researchers across five European institutions has made a case in The Lancet Infectious Diseases that artificial intelligence, paired with what epidemiologists call the One Health approach, could help catch dangerous pathogens before they spread widely among people. The argument is straightforward: most pandemics start in animals, but we cannot predict when or where a pathogen will jump to humans. By the time we notice it spreading, containment becomes brutally difficult—as the world learned with COVID-19. The question driving this work is whether better data integration and machine learning could shift that timeline, catching spillover events while they are still small.
Frank Møller Aarestrup, a professor at Denmark's DTU National Food Institute, and Marion Koopmans, from the Erasmus Medical Centre in the Netherlands, are among the authors. They emphasize that AI alone cannot prevent pandemics, but it can amplify what we already know how to do. The research team—which also includes experts from universities and health agencies in Hungary, Italy, and the United Kingdom—has spent years collaborating through the VEO consortium, a European initiative building data-driven tools to track emerging infectious diseases. Their argument rests on a simple observation: diseases like SARS-CoV-2, avian influenza, and mpox show us that spillovers are real, unpredictable, and increasingly common. Climate change, industrial animal farming, and human expansion into wild habitats all raise the odds that pathogens will jump species boundaries.
The researchers describe spillovers as sparks—most extinguish harmlessly, but some ignite fires that spread uncontrollably. The challenge is detecting those dangerous sparks early. This is where AI enters the picture. By analyzing datasets from climate records, land use patterns, animal production systems, transportation networks, population movements, and economic data, machine learning can surface patterns that would be invisible to human analysis alone. Aarestrup explains that AI can help identify geographic regions, specific animal species, wastewater systems, or human populations where surveillance should be intensified—essentially mapping the world's disease hotspots based on risk factors.
Once those hotspots are identified, a technique called metagenomic sequencing can be deployed as a broad surveillance net. This method analyzes genetic material from wastewater, air, food, or environmental samples to catalog the microorganisms present. When researchers sequence a sample, they often recover millions of genetic fragments. Most match known, harmless organisms, but thousands remain uncharacterized. This is where AI becomes essential: it can detect patterns in those unknown sequences and flag which ones might pose a threat. Koopmans notes that the genetic code itself contains clues about a virus's potential to infect humans, spread between people, and cause severe disease. AI-based protein models can now predict how mutations might alter viral structure and, by extension, transmissibility or disease severity—work that would take months or years through traditional methods.
The researchers also discuss emerging prototypes of AI "co-scientists"—systems capable of running entire research cycles, from generating hypotheses and reviewing literature to analyzing data and writing reports. Aarestrup envisions AI as a recognized competence at the research table, alongside human experts, offering analyses and suggestions that scientists can evaluate and build upon. But this vision comes with hard questions. Koopmans points out that researchers will need to learn new roles as supervisors and teachers, ensuring that AI-driven workflows produce trustworthy results and that humans can still recognize when the system makes mistakes. She acknowledges the challenge: as AI models advance, will we remain able to catch their errors? The authors conclude that artificial intelligence holds genuine promise for strengthening pandemic preparedness, but only if it remains a supplement to classical surveillance and research methods, not a replacement for them. The work ahead involves not just refining the technology, but rebuilding how scientists think about their own roles in an AI-augmented world.
Citas Notables
Artificial intelligence cannot by itself prevent pandemics, but the technology can be a powerful supplement to the knowledge and methods we already use.— Frank Møller Aarestrup, DTU National Food Institute
Once a disease starts spreading, it is very hard to bring under control. That is why it is crucial to detect new pathogens before they gain a foothold.— Marion Koopmans, Erasmus Medical Centre
Quieres la nota completa?
Lee el original en News-Medical.net ↗
Technology