Wastewater surveillance became a popular choice among public health officials looking to track rapid virus mutations and spread patterns during the COVID-19 pandemic. But what if there was a way to detect emerging viruses even faster — or to even sniff out new variants possibly before patients even realize they're ill?
A new UNLV-led study is moving that dream one step closer to reality by pairing wastewater sample surveillance with artificial intelligence. The results appear in the latest issue of the journal Nature Communications.
Lead author and UNLV neuroscience graduate student Xiaowei Zhuang developed an AI-driven algorithm that scans wastewater to detect budding influenza, RSV, mpox, measles, gonorrhea, Candida auris, or other pathogen variants — often before they're identified by clinical tests.
Scientists say being able to map virus emergence, mutation, and transmission faster with AI than with existing wastewater surveillance methods could significantly enhance public health officials' ability to roll out rapid, targeted interventions.
"Imagine identifying the next outbreak even before the first patient enters a clinic. This research shows how we can make this possible," said study co-author Edwin Oh, a professor with the Nevada Institute of Personalized Medicine at UNLV. "Through the use of AI we can determine how a pathogen is evolving without even testing a single human being."
While the study details how the team's AI method can separate overlapping signals in complex datasets, its real promise lies in on-the-ground impact. "The tool could especially be useful in improving disease surveillance in rural communities, empowering health workers in low-resource settings," said study co-author and Desert Research Institute research professor Duane Moser.
The research team tested its theory by analyzing nearly 3,700 wastewater samples collected from Southern Nevada wastewater treatment facilities between 2021 and 2023. They discovered that the AI-driven system could accurately identify unique signatures for different virus variants with as few as two to five samples, significantly earlier than existing methods.
Previous wastewater detection methods required prior knowledge of a variant's genetic makeup and relied heavily on clinical data from patients who had already been tested. Though those methods worked well, they were a more reactive approach — typically identifying new virus strains after they had already begun widely circulating in a community.
"Wastewater surveillance has enabled more timely and proactive public health responses through monitoring disease emergence and spread at a population level in real time," says Zhuang. "This new method enhances early outbreak detection to allow for identification of novel threats without prior knowledge or patient testing data, proactively detecting patterns from multiple wastewater samples and making this tool even more effective for public health surveillance moving forward."
Since 2021, four Las Vegas institutions – UNLV, the Southern Nevada Water Authority (SNWA), the Southern Nevada Health District, and the Desert Research Institute – have collaborated on a public wastewater surveillance dashboard to track emerging cases of COVID-19 and other viruses.
The Nature Communications AI study is one of more than 30 studies these organizations, along with the Cleveland Clinic Lou Ruvo Center for Brain Health, have collaborated on. And the researchers say it is among the first studies to employ an AI approach in enhancing wastewater intelligence.
"Wastewater surveillance has proven to be an effective tool for filling critical data gaps and understanding public health conditions within a community," said study co-author Daniel Gerrity, principal research microbiologist at SNWA. "The ongoing wastewater surveillance effort is a great example of how collaboration between SNWA, UNLV, and other partners can lead to positive impacts for the local community and beyond."
About The Study
" Early detection of emerging SARS-CoV-2 Variants from wastewater through genome sequencing and machine learning " was published July 8, 2025 in the journal Nature Communications.
