Optimising predictive models to prioritise viral discovery in zoonotic reservoirs

Abstract

Despite the global investment in One Health disease surveillance, it remains difficult and costly to identify and monitor the wildlife reservoirs of novel zoonotic viruses. Statistical models can guide sampling target prioritisation, but the predictions from any given model might be highly uncertain; moreover, systematic model validation is rare, and the drivers of model performance are consequently under-documented. Here, we use the bat hosts of betacoronaviruses as a case study for the data-driven process of comparing and validating predictive models of probable reservoir hosts. In early 2020, we generated an ensemble of eight statistical models that predicted host–virus associations and developed priority sampling recommendations for potential bat reservoirs of betacoronaviruses and bridge hosts for SARS-CoV-2. During a time frame of more than a year, we tracked the discovery of 47 new bat hosts of betacoronaviruses, validated the initial predictions, and dynamically updated our analytical pipeline. We found that ecological trait-based models performed well at predicting these novel hosts, whereas network methods consistently performed approximately as well or worse than expected at random. These findings illustrate the importance of ensemble modelling as a buffer against mixed-model quality and highlight the value of including host ecology in predictive models. Our revised models showed an improved performance compared with the initial ensemble, and predicted more than 400 bat species globally that could be undetected betacoronavirus hosts. We show, through systematic validation, that machine learning models can help to optimise wildlife sampling for undiscovered viruses and illustrates how such approaches are best implemented through a dynamic process of prediction, data collection, validation, and updating.

Full Genome Nobecovirus Sequences From Malagasy Fruit Bats Define a Unique Evolutionary History for This Coronavirus Clade

“…The fifth Betacoronavirus subgenus, Embecovirus, is primarily associated with rodents and bovids, though a few bat hosts have been documented (17,18). Since the emergence of SARS-CoV in 2002, there has been increasing interest in surveying potential hosts of coronaviruses and contributing new virus sequences to public databases, with most effort focused on sampling bats from Asia (19)(20)(21)(22)(23)(24)(25)(26)(27)(28), the continent of origin for both the SARS-CoV epidemic and the SARS-CoV-2 pandemic. Recently, more concerted efforts have arisen to survey the landscape of bat-borne coronaviruses in other regions of the world, including Africa and Europe (11,12,(29)(30)(31)(32)(33).…”

Serum proteomics identifies immune pathways and candidate biomarkers of coronavirus infection in wild vampire bats

“…Despite an increasing interest in bat–virus interactions, especially for CoVs given their human health relevance (Anthony et al, 2017b; Becker et al, 2022), we still have limited insights into the immune mechanisms involved in infection of bats (Ruiz-Aravena et al, 2021). Here, we used serum proteomics to broadly profile the immune phenotype of wild vampire bats in the presence of relatively common CoV infections.…”

“…Relationships between bats and CoVs specifically have been of particular interest for zoonotic risk assessment (Fischhoff et al, 2021; Becker et al, 2022). CoVs are RNA viruses found across mammals and birds, with at least seven known human viruses: two and five in the genera Alphacoronavirus and Betacoronavirus (Anthony et al, 2017b; Ye et al, 2020).…”

Predicting missing links in global host–parasite networks

“…An alternative is to treat hosts and parasites as two separate interacting classes that can be represented in a bipartite network (Albery et al, 2021). Bipartite network models can predict new links based solely on the structure of the observed network, or incorporate node‐level covariates, such as species traits (Becker et al, 2022; Dallas et al, 2017).…”

“…These may be derived from bipartite interaction data, but the derived networks effectively consider parasites as interchangeable. While sharing networks can identify host profiles for particular parasite groups or host species that promote parasite sharing across hosts, the ability to predict individual host–parasite interactions is limited (Becker et al, 2022). An alternative is to treat hosts and parasites as two separate interacting classes that can be represented in a bipartite network (Albery et al, 2021).…”