Abstract
Back and forth transmission of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) between humans and animals will establish wild reservoirs of virus that endanger long-term efforts to control COVID-19 in people and to protect vulnerable animal populations. Better targeting surveillance and laboratory experiments to validate zoonotic potential requires predicting high-risk host species. A major bottleneck to this effort is the few species with available sequences for angiotensin-converting enzyme 2 receptor, a key receptor required for viral cell entry. We overcome this bottleneck by combining species' ecological and biological traits with three-dimensional modelling of host-virus protein–protein interactions using machine learning. This approach enables predictions about the zoonotic capacity of SARS-CoV-2 for greater than 5000 mammals—an order of magnitude more species than previously possible. Our predictions are strongly corroborated by in vivo studies. The predicted zoonotic capacity and proximity to humans suggest enhanced transmission risk from several common mammals, and priority areas of geographic overlap between these species and global COVID-19 hotspots. With molecular data available for only a small fraction of potential animal hosts, linking data across biological scales offers a conceptual advance that may expand our predictive modelling capacity for zoonotic viruses with similarly unknown host ranges.
Highlights
The ongoing COVID-19 pandemic has surpassed 4.8 million deaths globally as of 1 October 2021 [1,2]
Like previous pandemics in recorded history, COVID-19 originated from the spillover of a zoonotic pathogen, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a betacoronavirus originating from an unknown animal host [3,4,5,6]
We combined structure-based models of viral binding with species-level data on biological and ecological traits to predict the capacity of mammal species to become zoonotic hosts of SARS-CoV-2
Summary
The ongoing COVID-19 pandemic has surpassed 4.8 million deaths globally as of 1 October 2021 [1,2]. We combine structural modelling of viral binding with machine learning of species ecological and biological traits to predict zoonotic capacity for SARS-CoV-2 across 5400 mammal species, expanding our predictive capacity by an order of magnitude (figure 2) This integrated approach enables predictions for the vast majority of species whose ACE2 sequences are currently unavailable by leveraging information from viral binding dynamics and biological traits. We identify a subset of species for which the threat of spillback infection appears greatest due to geographic overlaps and opportunities for contact with humans in areas of high SARS-CoV-2 prevalence globally These approaches underscore the utility of establishing interdisciplinary and iterative processes that join computational modelling, field surveillance and laboratory experiments to more efficiently quantify zoonotic risk [50], and better inform steps to prevent enzootic SARS-CoV-2 transmission and spread. Our analyses are based on the initial dominant SARS-CoV-2 variant in humans, but these methods can be readily adjusted to enable host range predictions for new variants as their hACE2-RBD crystal structures become available
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
