Zoonoses: A New Model for Identifying Viruses That Are Potentially Pathogenic to Humans

American researchers have developed a model that can predict not only whether a given zoonotic virus poses a risk of transmission to humans, but also the risk of human-to-human transmission. With an accuracy rate of over 84%, this model predicts human-to-human transmissibility with a high degree of precision. It has identified several animal viruses that represent a potential source of viral infectious disease in humans and may even be capable of spreading from person to person.   Emerging viral zoonoses pose a major threat to human health. However, developing a more proactive strategy for preventing epidemics requires new predictive tools. A recent study, published in PLOS ONE, shows that the human-to-human transmissibility of zoonotic viruses can be predicted with a high degree of accuracy using machine learning models trained on observed viral characteristics. It thus provides more refined potential targets for future disease surveillance and research. New pathogens emerging in humans most often originate from pathogens already circulating in animals. To understand how vectors and viruses interact with their hosts or with the environment, the researchers screened 224 viruses known to infect humans and, in particular, identified the viral characteristics associated with human-to-human transmission. A predictive machine learning model was applied to these data and was able to distinguish transmissible viruses from non-transmissible ones with 84% accuracy. When applied to all human viruses (both zoonotic and non-zoonotic), the model achieved even higher accuracy. The team uncovered several predictive traits in these viruses that spread among humans. The detection of the virus in the liver, the central nervous system, and the upper and lower respiratory tracts was associated with an increased likelihood of human-to-human transmission. Similarly, zoonoses transmitted by monkeys were more likely to be transmissible between humans than other zoonotic viruses found in rodents and birds, among others. The absence of a lipid envelope, the small size of viral particles, and limited genome segmentation were also associated with human transmissibility. Using the main model, the viruses were ranked according to their probability of human-to-human transmission. Of the 85 human viruses for which it is currently unknown whether they can circulate from person to person, 47 showed a high probability of secondary human transmission: Carnivore amdoparvovirus 1, Hendra virus, cardiovirus A, Rosavirus A, the human T-lymphotropic viruses HTLV-3 and HTLV-4, and simian immunodeficiency virus. Future epidemiological studies on these pathogens should now focus on identifying potential human-to-human transmission, as underreporting of cases and misdiagnoses often allow these viruses to spread undetected. The ability to predict which viruses are capable of spreading among humans and are thus likely to cause human epidemics has direct practical implications. Public health agencies could use this model as a tool to prioritize disease surveillance and characterize the epidemic risk posed by new viral pathogens.