If the COVID-19 epidemic has taught us anything, it’s that for all our scientific and medical advances, the human race is still very vulnerable to novel viruses. These novel viruses lurk in wild animal populations and can spread to people who come in contact with infected animals. Alternatively, wild animal viruses can infect domestic animals who may then pass it on to their human contacts. Lastly, some animal viruses are transmitted to people via blood-sucking insects such as ticks, fleas, or mosquitos. When an animal virus infects a human it is called a zoonotic infection.
Some well-known zoonotic diseases include rabies, Zika, and West Nile virus, all of which occur sporadically when a human gets infected with the causative animal virus. However, we know what animals carry these viruses and how the viruses get transmitted to humans, so we can take preventative measures. What we don’t know is how many novel viruses exist in all the animal species of the world and what fraction of these viruses can infect and cause disease in humans. This problem is exemplified just by comparing humans to primates. A 2015 publication identified 274 viral species infecting humans, while we currently only know an average of 7 viruses infecting each of the hundreds of primate species. It is highly unlikely that humans have 40 times more viruses than other primates, suggesting that there is an extensive reservoir of primate viruses that just haven’t been discovered yet. Extend this to all the animals beyond primates and the scope of the problem becomes enormous. Even if we just consider mammals, most estimates suggest that we’ve identified less than 1% of the viruses that infect this subgroup of animals. Some of these known viruses can infect a broad range of species whereas others have a very narrow host range, and we would expect a similar distribution for the unknown viruses. Host range is an important property because animal viruses with broad host ranges are more likely capable of spilling over from their natural host into a foreign host such as a domestic animal or a human. Knowing or anticipating what hosts a virus can infect would be helpful for developing control and containment procedures for new outbreaks.
A British research group recently developed a machine-learning algorithm to predict virus-mammal associations. Their results, published in Nature Communications, used 6,331 recognized associations between 1896 viruses and 1436 mammals (an average of 3.3 hosts per virus) as the training set for their algorithm. Once trained, the algorithm was used to predict likely virus-host interactions that had not yet been observed. While such predictions must always be taken with some reservation, their results revealed an intriguing trend. The algorithm suggested that there were over 20,000 more potential interactions among this training set of viruses and hosts. This translates to each virus being able to infect an average of about 14 different host species. This result suggests that our current knowledge significantly underestimates the number of hosts that each virus can infect by a factor of roughly 4. Furthermore, the predictions can even be targeted to the level of individual viruses or hosts. For example, the algorithm identified 44 possible novel associations between bats and Filoviruses, the group that includes Ebola and other viruses causing hemorrhagic fevers. Similarly, applying this tool to other viruses suggested numerous potential virus interactions with hosts that have not yet been detected. Overall, the results in this paper suggest that our current assessment of viral host range significantly underestimates the true scope of possible diverse host infections. Hopefully, this and other predictive algorithms can help direct scientists to study those mammalian species and viral families where our knowledge is most limited and our greatest risk of potential zoonotic outbreaks exists. We’ve seen and experienced the medical, social, and economic devastation caused by SARS-CoV-2, and our best defense against future novel viral pandemics is aggressive research to identify and limit potential new threats.