Can AI help spot ‘healthy’ viruses in our bodies? UB researchers probe the human virome

BUFFALO, N.Y. ­–  When it comes to viruses, it’s long been the case of the naughty kid getting all the attention. For decades, scientists have focused primarily on the viruses that cause human illness, including influenza, chicken pox and COVID-19.

Much less is understood about the numerous viruses in our bodies that are not associated with any type of disease.

Researchers representing three schools and a center within the University at Buffalo are studying these very viruses and how combining laboratory methods and artificial intelligence (AI) can help identify them.

Their project “Hiding in plain sight: Integrating AI with targeted bench methods to discover and characterize viruses in the human body,” recently received a $423,366 grant from the National Institutes of Health. It’s a four-year renewable grant totaling $1.6 million.

“Healthy humans are full of trillions of viruses. Studying them is challenging because they are far more diverse than bacteria, and it can be hard to see ‘who’ is there and discern what they are doing in the body,” says Kathryn Kauffman, PhD, assistant professor in the Department of Oral Biology in the School of Dental Medicine.

She is co-leading this multiple principal investigator project with Libusha Kelly, PhD, associate professor of microbiology and immunology, and systems and computational biology at Albert Einstein College of Medicine.

“Libusha and her team are developing new ways of discovering viral sequences and what viral proteins are doing,” Kauffman says. “Similarly to how large language models learn about human language and underpin generative AI, protein language models (PLMs) can learn about biological ‘words’ like viral protein sequences. Libusha’s team is developing innovative new tools for understanding viruses in the human body by building on these foundational PLMs.”

Along with Kauffman, the UB team includes Yinyin Ye, PhD, assistant professor in the Department of Civil, Structural and Environmental Engineering in the School of Engineering and Applied Sciences, and Jun Qu, PhD, professor in the Department of Pharmaceutical Sciences in the School of Pharmacy and Pharmaceutical Sciences and the New York State Center of Excellence in Bioinformatics and Life Sciences (CBLS).

Ye will bring her expertise in concentrating and purifying viruses from wastewater and viral proteomics. Meanwhile, Qu will share his experience using a state-of-the art mass spectrometer, which is located in the CBLS’s Proteomics and Bioanalysis Core.

“The rapid evolution of mass spectrometry technologies and bioinformatics has dramatically expanded our ability to identify and characterize proteins on a large scale,” Qu says. “In particular, next-generation mass spectrometry platforms, such as the Orbitrap Astral, significantly improve sensitivity, selectivity and analytical throughput. These advances enable the detection of many low-abundance proteins that were previously difficult or impossible to observe, including viral proteins that may play important roles in the human virome being investigated in this project.”

“By integrating cutting-edge proteomics with AI-driven bioinformatics approaches, we hope to uncover previously hidden viruses and gain new insights into their biological functions in healthy humans,” Qu explains.

Their research is part of the NIH Human Virome Program (HVP), a consortium involving more than 300 researchers across the country who are working to improve scientists’ understanding of the role viruses play in the healthy human body. The virome is the entire collection of viruses living within the human body.

The HVP has awarded more than $100 million to multiple institutions, including teams at five Virome Characterization Centers across the U.S.: the University of Pennsylvania, Broad Institute of MIT and Harvard, Vanderbilt University, Stanford University, and the University of California, Los Angeles.

There are also teams focusing on the areas of functional studies, data coordination across the consortium and development of new tools and methods. UB is focusing on the third category, with the goal of developing methods for uncovering and identifying viruses that are usually missed in virome studies.

To ferret out the hidden viruses, the team will collect viral samples and process them in a way that removes a dominant group of viruses that normally “swamp out” the signal from other viruses.

“Through multiple steps, we hope to concentrate viruses that were previously ignored, and we’ll do DNA and RNA sequencing and protein characterization,” Ye explains.

During her doctoral and postdoctoral work at the MIT-Woods Hole Oceanographic Institution, Kauffman helped discover a new kind of virus.

“It revealed features that we are building off of with this project,” she says. “By the end of this project, we may say, ‘Here are hundreds of new types of viruses we hadn’t been able to see before.’ Now we can look for their sequences in all the human sample datasets generated across the HVP consortium.”

The UB team will use a source of viruses that is plentiful and rich in diverse viruses present in healthy humans — wastewater. Using wastewater accelerates the development of tools for general study of the human virome, and it also holds potential future benefit for public health studies.

“From the engineering side, if we better understand the viruses associated with healthy humans,” Ye says, “we can apply this knowledge and leverage wastewater-based epidemiology to monitor for anomaly signals and provide early warning for community-wide disease outbreaks.”

Kauffman notes that the UB and Einstein teams are in regular conversations with others across the HVP consortium.

“A central goal of the HVP is to gain a fuller picture of viruses in the ‘healthy’ human virome,” Kauffman says. “And we have brought together an exceptional team here at UB and Einstein to be a part of making that happen.”