Wildlife susceptibility to infectious diseases at global scales, is a study published in the Proceedings of the National Academy of Sciences of the United States of America. In their retrospective, the researchers analyze: "Disease transmission prediction across wildlife is crucial for risk assessment of emerging infectious diseases.
Susceptibility of host species to pathogens is influenced by the geographic, environmental, and phylogenetic context of the specific system under study. We used machine learning to analyze how such variables influence pathogen incidence for multihost pathogen assemblages, including one of direct transmission (coronaviruses and bats) and two vector-borne systems.
Here we show that this methodology is able to provide reliable global spatial susceptibility predictions for the studied host-pathogen systems, even when using a small amount of incidence information. We found that avian malaria was mostly affected by environmental factors and by an interaction between phylogeny and geography, and WNV susceptibility was mostly influenced by phylogeny and by the interaction between geographic and environmental distances, whereas coronavirus susceptibility was mostly affected by geography.
This approach will help to direct surveillance and field efforts providing cost-effective decisions on where to invest limited resources. Geography was the most important factor determining bat coronavirus susceptibility, which is modulated by the interaction of geography with host phylogeny, and is supported by a study of viral communities in bats and rodents.
We obtained geographic distribution information for birds and mammals from the IUCN polygons. We first calculated the centroid of the polygon with the largest area for each species to get the geographic information. Subsequently, we calculated the geographical distance between each centroid to generate a geographic distance matrix.
Previous studies have investigated how environmental, phylogenetic, and geographic variables determine pathogen infection, particularly for human zoonosis. Yet, none of those previous studies have provided a methodology that can be applied to a broad array of host–pathogen systems.
Here, we provide a machine learning approach that can integrate different explanatory variables and be applied to any multihost–multipathogen system. Our results agree with the known ecology of each analyzed system and provide a tool that can help discovering potential host species and novel geographical hot spots for a pathogen.
Thus, it can help guiding sampling decisions in terms of both host species and geographical locations. Finally, this tool can be applied at different spatial scales with few incidence data."