Where coronaviruses hide, where novel strains are generated, and how they get to us: Predicting reservoirs, recombination, and geographical hotspots

Novel pathogenic coronaviruses, including SARS-CoV and SARS-CoV-2, arise by genetic recombining of two different coronavirus strains co-infecting an animal host. These viruses circulate in reservoir animal populations before spillover to humans. Understanding, monitoring, and mitigating both recombination and spillover requires identifying hosts that are susceptible to each coronavirus and hosts susceptible to multiple coronavirus strains (termed recombination hosts). However, the majority of coronavirus-host associations, and therefore reservoirs and recombination hosts, remain unidentified. This has led to an underappreciation of the potential scale of novel coronavirus generation and spillover. Here, we aim to predict all host species which act as SARS-CoV-2 reservoirs and recombination hosts (WP1), by expanding our tried-and-tested machine-learning framework to include avian hosts. This will enable monitoring of SARS-CoV-2 reservoirs during the pandemic, and hosts in which SARS-CoV-2 could recombine to generate novel pathogenic viruses. Geographical overlap of host species is a key predictor of between-species viral sharing. By constructing a species-level ecological contact network and integrating it with our framework we will further refine our predictions. This will enable us to identify geographical hotspots of coronaviruses recombination (WP2), and therefore enable specific spatially-targeted surveillance and mitigation efforts. Many coronavirus hosts interact with humans, either naturally by e.g. geographic/habitat overlap, or are used by humans as e.g. pets/food. Using geographical data from WP2 and host species utilisation data from open-access sources, we will (WP3) estimate the in situ likelihood of spillover from hosts identified in WP1. This will inform policy makers of species and hotspots for spillover mitigation efforts.