Innovative Model Maps Where Poxviruses May Strike Next

The legacy of smallpox eradication is a double-edged sword. While it marked one of the greatest achievements in public health, the discontinuation of the smallpox vaccine has left much of the global population vulnerable to its viral relatives—the orthopoxviruses. These include mpox, cowpox, and emerging threats like borealpox virus, all of which continue to circulate in wildlife and occasionally spill into human populations. A new study, published in Communications Biology, suggests that the key to predicting where and in which animals such spillovers might occur lies in the viruses' genomes.

Led by researchers from Washington State University, the study integrates viral genomic data with mammalian ecological traits to model which species are likely to harbor orthopoxviruses (OPVs). Their approach marks a significant departure from traditional models, which have relied almost exclusively on host traits like geographic range, body size, or evolutionary lineage. By incorporating genetic features—particularly viral accessory genes known to influence immune evasion and host compatibility—the team substantially improved the accuracy of their predictions.

Using machine learning algorithms called boosted regression trees (BRTs), the researchers compared models that used only host traits against those that combined host traits with viral genomic features. The hybrid model not only performed better overall but also revealed telling patterns: for instance, felids (such as wild cats) were consistently predicted to be likely OPV hosts, while rodents and lagomorphs (like rabbits) were less so. When zooming in on specific virus-host pairings, the influence of the viral genome became even more apparent. Genes linked to host immune modulation, membrane proteins, and cytokine inhibitors were strong indicators of which species a virus might infect.

The practical implications of this model are striking. The team mapped their predictions geographically and found alarming overlaps between areas rich in potential OPV hosts and regions with historically low smallpox vaccination rates. Central and East Africa, parts of Southeast Asia, and the Amazon basin emerged as critical zones of vulnerability. These areas not only harbor a high density of likely OPV reservoir species but also house populations with little to no immunity to poxviruses.

One particularly revealing case study involved borealpox virus—a newly identified orthopoxvirus isolated in humans in Alaska. Using a more permissive model threshold, the researchers identified the genus Mus (common mice) as a likely reservoir. This prediction aligned with known viral recombination events between borealpox and ectromelia virus, a poxvirus that infects mice. Such genomic clues offer more than predictive power—they provide molecular hypotheses that can guide targeted surveillance.

Yet the study does not shy away from its limitations. The authors note the challenges of incomplete sampling, data sparsity for some OPV genomes, and reliance on pseudoabsences (species assumed not to be hosts due to lack of evidence). They also acknowledge that machine learning models, especially those using principal component analysis on genomic data, can be difficult to interpret biologically. Nevertheless, they argue these models can prioritize candidates for field and lab investigations in ways that purely ecological approaches cannot.

As mpox continues to appear in regions outside its historic range and human-animal viral exchanges become more frequent, understanding where the next outbreak might spark is no longer just academic. This study provides a vital tool for public health officials, ecologists, and virologists alike—one that reads between the lines of the genetic code to anticipate the next zoonotic leap.

Source:

Tseng, K.K., Koehler, H., Becker, D.J. et al. Viral genomic features predict Orthopoxvirus reservoir hosts. Commun Biol8, 309 (2025). https://doi.org/10.1038/s42003-025-07746-0