Predicting zoonotic potential of viruses: where are we?

The prospect of identifying high-risk viruses and designing interventions to pre-empt their emergence into human populations is enticing, but controversial, particularly when used to justify large-scale virus discovery initiatives. We review the current state of these efforts, identifying three broad classes of predictive models that have differences in data inputs that define their potential utility for triaging newly discovered viruses for further investigation. Prospects for model predictions of public health risk to guide preparedness depend not only on computational improvements to algorithms, but also on more efficient data generation in laboratory, field and clinical settings. Beyond public health applications, efforts to predict zoonoses provide unique research value by creating generalisable understanding of the ecological and evolutionary factors that promote viral emergence.

Network embedding unveils the hidden interactions in the mammalian virome.

Predicting host-virus interactions is fundamentally a network science problem. We develop a method for bipartite network prediction that combines a recommender system (linear filtering) with an imputation algorithm based on low-rank graph embedding. We test this method by applying it to a global database of mammal-virus interactions and thus show that it makes biologically plausible predictions that are robust to data biases. We find that the mammalian virome is under-characterized anywhere in the world. We suggest that future virus discovery efforts could prioritize the Amazon Basin (for its unique coevolutionary assemblages) and sub-Saharan Africa (for its poorly characterized zoonotic reservoirs). Graph embedding of the imputed network improves predictions of human infection from viral genome features, providing a shortlist of priorities for laboratory studies and surveillance. Overall, our study indicates that the global structure of the mammal-virus network contains a large amount of information that is recoverable, and this provides new insights into fundamental biology and disease emergence.

Variation in the ACE2 receptor has limited utility for SARS-CoV-2 host prediction.

Transmission of SARS-CoV-2 from humans to other species threatens wildlife conservation and may create novel sources of viral diversity for future zoonotic transmission. A variety of computational heuristics have been developed to pre-emptively identify susceptible host species based on variation in the angiotensin-converting enzyme 2 (ACE2) receptor used for viral entry. However, the predictive performance of these heuristics remains unknown. Using a newly compiled database of 96 species, we show that, while variation in ACE2 can be used by machine learning models to accurately predict animal susceptibility to sarbecoviruses (accuracy = 80.2%, binomial confidence interval [CI]: 70.8-87.6%), the sites informing predictions have no known involvement in virus binding and instead recapitulate host phylogeny. Models trained on host phylogeny alone performed equally well (accuracy = 84.4%, CI: 75.5-91.0%) and at a level equivalent to retrospective assessments of accuracy for previously published models. These results suggest that the predictive power of ACE2-based models derives from strong correlations with host phylogeny rather than processes which can be mechanistically linked to infection biology. Further, biased availability of ACE2 sequences misleads projections of the number and geographic distribution of at-risk species. Models based on host phylogeny reduce this bias, but identify a very large number of susceptible species, implying that model predictions must be combined with local knowledge of exposure risk to practically guide surveillance. Identifying barriers to viral infection or onward transmission beyond receptor binding and incorporating data which are independent of host phylogeny will be necessary to manage the ongoing risk of establishment of novel animal reservoirs of SARS-CoV-2.

The COVID-19 pandemic affects humans, but also many of the animals we interact with. So far, humans have transmitted the SARS-CoV-2 virus to pet dogs and cats, a wide range of zoo animals, and even wildlife. Transmission of SARS-CoV-2 from humans to animals can lead to outbreaks amongst certain species, which can endanger animal populations and create new sources of human infections. Thus, careful monitoring of animal infections may help protect both animals and humans. Identifying which animals are susceptible to SARS-CoV-2 would help scientists monitor these species for outbreaks and viral circulation. Unfortunately, testing whether SARS-CoV-2 can infect different species in the laboratory is both time-consuming and expensive. To overcome this obstacle, researchers have used computational methods and existing data about the structure and genetic sequences of ACE2 receptors ­ the proteins on the cell surface that SARS-CoV-2 uses to enter the cell ­ to predict SARS-COV-2 susceptibility in different species. However, it remained unclear how accurate this approach was at predicting susceptibility in different animals, or whether their correct predictions indicated causal links between ACE2 variability and SARS-CoV-2 susceptibility. To assess the usefulness of this approach, Mollentze et al. started by using data on the ACE2 receptors from 96 different species and building a machine learning model to predict how susceptible those species might be to SARS-CoV-2. The susceptibility of these species had either been observed in natural infections ­ in zoos, for example ­ or had been assessed in the laboratory, so Mollentze et al. were able to use this information to determine how good both their model and previous approaches based on the sequence of ACE2 receptors were. The results showed that while the model was quite accurate (it correctly predicted susceptibility to SARS-CoV-2 about 80% of the time), its predictions were based on regions of the ACE2 receptors that were not known to interact with the virus. Instead, the regions that the machine learning model relied on were ones that tend to vary more the more distantly related two species are. This indicates that existing computational approaches are likely not relying on information about how ACE2 receptors interact with SARS-CoV-2 to predict susceptibility. Instead, they are simply using information on how closely related the different animal species are, which is much easier to source than data about ACE2 receptors. Indeed, the sequences of the ACE2 receptors in many species are unknown and the species for which this information is available come only from a few geographic areas. Mollentze et al. also showed that limiting the predictions about susceptibility to these species could mislead scientists when deciding which species and geographic areas to surveil for possible viral circulation. Instead, it may be more effective and cost-efficient to use animal relatedness to predict susceptibility to SARS-CoV-2. This makes it possible to make predictions for nearly all mammals, while being just as accurate as models based on ACE2 receptor data. However, Mollentze et al. point out that this approach would still fail to narrow down the number of animals that need to be monitored enough for it to be practical. Considering additional factors like how often the animals interact with humans or how prone they are to transmit the virus among themselves may help narrow it down more. Further research is therefore needed to identify the best multifactor approaches to identifying which animal populations should be monitored.

The apparent interferon resistance of transmitted HIV-1 is possibly a consequence of enhanced replicative fitness.

HIV-1 transmission via sexual exposure is an inefficient process. When transmission does occur, newly infected individuals are colonized by the descendants of either a single virion or a very small number of establishing virions. These transmitted founder (TF) viruses are more interferon (IFN)-resistant than chronic control (CC) viruses present 6 months after transmission. To identify the specific molecular defences that make CC viruses more susceptible to the IFN-induced 'antiviral state', we established a single pair of fluorescent TF and CC viruses and used arrayed interferon-stimulated gene (ISG) expression screening to identify candidate antiviral effectors. However, we observed a relatively uniform ISG resistance of transmitted HIV-1, and this directed us to investigate possible underlying mechanisms. Simple simulations, where we varied a single parameter, illustrated that reduced growth rate could possibly underly apparent interferon sensitivity. To examine this possibility, we closely monitored in vitro propagation of a model TF/CC pair (closely matched in replicative fitness) over a targeted range of IFN concentrations. Fitting standard four-parameter logistic growth models, in which experimental variables were regressed against growth rate and carrying capacity, to our in vitro growth curves, further highlighted that small differences in replicative growth rates could recapitulate our in vitro observations. We reasoned that if growth rate underlies apparent interferon resistance, transmitted HIV-1 would be similarly resistant to any growth rate inhibitor. Accordingly, we show that two transmitted founder HIV-1 viruses are relatively resistant to antiretroviral drugs, while their matched chronic control viruses were more sensitive. We propose that, when present, the apparent IFN resistance of transmitted HIV-1 could possibly be explained by enhanced replicative fitness, as opposed to specific resistance to individual IFN-induced defences. However, further work is required to establish how generalisable this mechanism of relative IFN resistance might be.

Ecological determinants of rabies virus dynamics in vampire bats and spillover to livestock.

The pathogen transmission dynamics in bat reservoirs underpin efforts to reduce risks to human health and enhance bat conservation, but are notoriously challenging to resolve. For vampire bat rabies, the geographical scale of enzootic cycles, whether environmental factors modulate baseline risk, and how within-host processes affect population-level dynamics remain unresolved. We studied patterns of rabies exposure using an 11-year, spatially replicated sero-survey of 3709 Peruvian vampire bats and co-occurring outbreaks in livestock. Seroprevalence was correlated among nearby sites but fluctuated asynchronously at larger distances. A generalized additive mixed model confirmed spatially compartmentalized transmission cycles, but no effects of bat demography or environmental context on seroprevalence. Among 427 recaptured bats, we observed long-term survival following rabies exposure and antibody waning, supporting hypotheses that immunological mechanisms influence viral maintenance. Finally, seroprevalence in bats was only weakly correlated with outbreaks in livestock, reinforcing the challenge of spillover prediction even with extensive data. Together our results suggest that rabies maintenance requires transmission among multiple, nearby bat colonies which may be facilitated by waning of protective immunity. However, the likelihood of incursions and dynamics of transmission within bat colonies appear largely independent of bat ecology. The implications of these results for spillover anticipation and controlling transmission at the source are discussed.

Bats host the most virulent-but not the most dangerous-zoonotic viruses.

SignificanceThe clear need to mitigate zoonotic risk has fueled increased viral discovery in specific reservoir host taxa. We show that a combination of viral and reservoir traits can predict zoonotic virus virulence and transmissibility in humans, supporting the hypothesis that bats harbor exceptionally virulent zoonoses. However, pandemic prevention requires thinking beyond zoonotic capacity, virulence, and transmissibility to consider collective "burden" on human health. For this, viral discovery targeting specific reservoirs may be inefficient as death burden correlates with viral, not reservoir, traits, and depends on context-specific epidemiological dynamics across and beyond the human-animal interface. These findings suggest that longitudinal studies of viral dynamics in reservoir and spillover host populations may offer the most effective strategy for mitigating zoonotic risk.

Mammal virus diversity estimates are unstable due to accelerating discovery effort.

Host-virus association data underpin research into the distribution and eco-evolutionary correlates of viral diversity and zoonotic risk across host species. However, current knowledge of the wildlife virome is inherently constrained by historical discovery effort, and there are concerns that the reliability of ecological inference from host-virus data may be undermined by taxonomic and geographical sampling biases. Here, we evaluate whether current estimates of host-level viral diversity in wild mammals are stable enough to be considered biologically meaningful, by analysing a comprehensive dataset of discovery dates of 6571 unique mammal host-virus associations between 1930 and 2018. We show that virus discovery rates in mammal hosts are either constant or accelerating, with little evidence of declines towards viral richness asymptotes, even in highly sampled hosts. Consequently, inference of relative viral richness across host species has been unstable over time, particularly in bats, where intensified surveillance since the early 2000s caused a rapid rearrangement of species' ranked viral richness. Our results illustrate that comparative inference of host-level virus diversity across mammals is highly sensitive to even short-term changes in sampling effort. We advise caution to avoid overinterpreting patterns in current data, since it is feasible that an analysis conducted today could draw quite different conclusions than one conducted only a decade ago.

The science of the host-virus network.

Better methods to predict and prevent the emergence of zoonotic viruses could support future efforts to reduce the risk of epidemics. We propose a network science framework for understanding and predicting human and animal susceptibility to viral infections. Related approaches have so far helped to identify basic biological rules that govern cross-species transmission and structure the global virome. We highlight ways to make modelling both accurate and actionable, and discuss the barriers that prevent researchers from translating viral ecology into public health policies that could prevent future pandemics.

The antiviral state has shaped the CpG composition of the vertebrate interferome to avoid self-targeting.

Antiviral defenses can sense viral RNAs and mediate their destruction. This presents a challenge for host cells since they must destroy viral RNAs while sparing the host mRNAs that encode antiviral effectors. Here, we show that highly upregulated interferon-stimulated genes (ISGs), which encode antiviral proteins, have distinctive nucleotide compositions. We propose that self-targeting by antiviral effectors has selected for ISG transcripts that occupy a less self-targeted sequence space. Following interferon (IFN) stimulation, the CpG-targeting antiviral effector zinc-finger antiviral protein (ZAP) reduces the mRNA abundance of multiple host transcripts, providing a mechanistic explanation for the repression of many (but not all) interferon-repressed genes (IRGs). Notably, IRGs tend to be relatively CpG rich. In contrast, highly upregulated ISGs tend to be strongly CpG suppressed. Thus, ZAP is an example of an effector that has not only selected compositional biases in viral genomes but also appears to have notably shaped the composition of host transcripts in the vertebrate interferome.

The future of zoonotic risk prediction.

In the light of the urgency raised by the COVID-19 pandemic, global investment in wildlife virology is likely to increase, and new surveillance programmes will identify hundreds of novel viruses that might someday pose a threat to humans. To support the extensive task of laboratory characterization, scientists may increasingly rely on data-driven rubrics or machine learning models that learn from known zoonoses to identify which animal pathogens could someday pose a threat to global health. We synthesize the findings of an interdisciplinary workshop on zoonotic risk technologies to answer the following questions. What are the prerequisites, in terms of open data, equity and interdisciplinary collaboration, to the development and application of those tools? What effect could the technology have on global health? Who would control that technology, who would have access to it and who would benefit from it? Would it improve pandemic prevention? Could it create new challenges? This article is part of the theme issue 'Infectious disease macroecology: parasite diversity and dynamics across the globe'.

Identifying and prioritizing potential human-infecting viruses from their genome sequences.

Determining which animal viruses may be capable of infecting humans is currently intractable at the time of their discovery, precluding prioritization of high-risk viruses for early investigation and outbreak preparedness. Given the increasing use of genomics in virus discovery and the otherwise sparse knowledge of the biology of newly discovered viruses, we developed machine learning models that identify candidate zoonoses solely using signatures of host range encoded in viral genomes. Within a dataset of 861 viral species with known zoonotic status, our approach outperformed models based on the phylogenetic relatedness of viruses to known human-infecting viruses (area under the receiver operating characteristic curve [AUC] = 0.773), distinguishing high-risk viruses within families that contain a minority of human-infecting species and identifying putatively undetected or so far unrealized zoonoses. Analyses of the underpinnings of model predictions suggested the existence of generalizable features of viral genomes that are independent of virus taxonomic relationships and that may preadapt viruses to infect humans. Our model reduced a second set of 645 animal-associated viruses that were excluded from training to 272 high and 41 very high-risk candidate zoonoses and showed significantly elevated predicted zoonotic risk in viruses from nonhuman primates, but not other mammalian or avian host groups. A second application showed that our models could have identified Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) as a relatively high-risk coronavirus strain and that this prediction required no prior knowledge of zoonotic Severe Acute Respiratory Syndrome (SARS)-related coronaviruses. Genome-based zoonotic risk assessment provides a rapid, low-cost approach to enable evidence-driven virus surveillance and increases the feasibility of downstream biological and ecological characterization of viruses.

Characterizing and Evaluating the Zoonotic Potential of Novel Viruses Discovered in Vampire Bats.

The contemporary surge in metagenomic sequencing has transformed knowledge of viral diversity in wildlife. However, evaluating which newly discovered viruses pose sufficient risk of infecting humans to merit detailed laboratory characterization and surveillance remains largely speculative. Machine learning algorithms have been developed to address this imbalance by ranking the relative likelihood of human infection based on viral genome sequences, but are not yet routinely applied to viruses at the time of their discovery. Here, we characterized viral genomes detected through metagenomic sequencing of feces and saliva from common vampire bats (Desmodus rotundus) and used these data as a case study in evaluating zoonotic potential using molecular sequencing data. Of 58 detected viral families, including 17 which infect mammals, the only known zoonosis detected was rabies virus; however, additional genomes were detected from the families Hepeviridae, Coronaviridae, Reoviridae, Astroviridae and Picornaviridae, all of which contain human-infecting species. In phylogenetic analyses, novel vampire bat viruses most frequently grouped with other bat viruses that are not currently known to infect humans. In agreement, machine learning models built from only phylogenetic information ranked all novel viruses similarly, yielding little insight into zoonotic potential. In contrast, genome composition-based machine learning models estimated different levels of zoonotic potential, even for closely related viruses, categorizing one out of four detected hepeviruses and two out of three picornaviruses as having high priority for further research. We highlight the value of evaluating zoonotic potential beyond ad hoc consideration of phylogeny and provide surveillance recommendations for novel viruses in a wildlife host which has frequent contact with humans and domestic animals.

Virulence mismatches in index hosts shape the outcomes of cross-species transmission.

Whether a pathogen entering a new host species results in a single infection or in onward transmission, and potentially an outbreak, depends upon the progression of infection in the index case. Although index infections are rarely observable in nature, experimental inoculations of pathogens into novel host species provide a rich and largely unexploited data source for meta-analyses to identify the host and pathogen determinants of variability in infection outcomes. We analyzed the progressions of 514 experimental cross-species inoculations of rabies virus, a widespread zoonosis which in nature exhibits both dead-end infections and varying levels of sustained transmission in novel hosts. Inoculations originating from bats rather than carnivores, and from warmer- to cooler-bodied species caused infections with shorter incubation periods that were associated with diminished virus excretion. Inoculations between distantly related hosts tended to result in shorter clinical disease periods, which are also expected to impede onward transmission. All effects were modulated by infection dose. Taken together, these results suggest that as host species become more dissimilar, increased virulence might act as a limiting factor preventing onward transmission. These results can explain observed constraints on rabies virus host shifts, describe a previously unrecognized role of host body temperature, and provide a potential explanation for host shifts being less likely between genetically distant species. More generally, our study highlights meta-analyses of experimental infections as a tractable approach to quantify the complex interactions between virus, reservoir, and novel host that shape the outcome of cross-species transmission.

Viral zoonotic risk is homogenous among taxonomic orders of mammalian and avian reservoir hosts.

The notion that certain animal groups disproportionately maintain and transmit viruses to humans due to broad-scale differences in ecology, life history, and physiology currently influences global health surveillance and research in disease ecology, virology, and immunology. To directly test whether such "special reservoirs" of zoonoses exist, we used literature searches to construct the largest existing dataset of virus-reservoir relationships, consisting of the avian and mammalian reservoir hosts of 415 RNA and DNA viruses along with their histories of human infection. Reservoir host effects on the propensity of viruses to have been reported as infecting humans were rare and when present were restricted to one or two viral families. The data instead support a largely host-neutral explanation for the distribution of human-infecting viruses across the animal orders studied. After controlling for higher baseline viral richness in mammals versus birds, the observed number of zoonoses per animal order increased as a function of their species richness. Animal orders of established importance as zoonotic reservoirs including bats and rodents were unexceptional, maintaining numbers of zoonoses that closely matched expectations for mammalian groups of their size. Our findings show that variation in the frequency of zoonoses among animal orders can be explained without invoking special ecological or immunological relationships between hosts and viruses, pointing to a need to reconsider current approaches aimed at finding and predicting novel zoonoses.

Rabies Virus.

This infographic describes the transmission cycle of rabies virus in domestic dogs and the necessity of a One Health approach, integrating medical and veterinary interventions, to control and eliminate human rabies deaths. Rabies virus (RABV) causes an acute, fatal neurological infection in humans and other mammals, transmitted through the saliva of rabid animals via a bite or scratch. From the site of infection the virus travels along neurons to the central nervous system (CNS), where viral replication leads to symptoms and systemic spread. Once symptomatic, the disease is nearly 100% fatal. However, the disease is 100% vaccine-preventable through the prompt administration of human postexposure prophylaxis (PEP) and vaccination of animal reservoirs. While RABV has a broad host range, domestic dogs cause over 99% of all human cases, killing 59000 people every year. Human PEP is costly (US$11-150 per dose) and often difficult to obtain. Dog vaccination is a considerably more cost-effective and feasible method to reduce the incidence of human rabies. With this in mind, the World Health Organisation (WHO) and partners have set a target for the global elimination of dog-mediated human rabies, through control of the disease in dogs, by 2030.

The role of viral evolution in rabies host shifts and emergence.

Despite its ability to infect all mammals, Rabies virus persists in numerous species-specific cycles that rarely sustain transmission in alternative species. The determinants of these species-associations and the adaptive significance of genetic divergence between host-associated viruses are poorly understood. One explanation is that epidemiological separation between reservoirs causes neutral genetic differentiation. Indeed, recent studies attributed host shifts to ecological factors and selection of 'preadapted' viral variants from the existing viral community. However, phenotypic differences between isolates and broad scale comparative and molecular evolutionary analyses indicate multiple barriers that Rabies virus must overcome through adaptation. This review assesses various lines of evidence and proposes a synthetic hypothesis for the respective roles of ecology and evolution in Rabies virus host shifts.

A Bayesian approach for inferring the dynamics of partially observed endemic infectious diseases from space-time-genetic data.

We describe a statistical framework for reconstructing the sequence of transmission events between observed cases of an endemic infectious disease using genetic, temporal and spatial information. Previous approaches to reconstructing transmission trees have assumed all infections in the study area originated from a single introduction and that a large fraction of cases were observed. There are as yet no approaches appropriate for endemic situations in which a disease is already well established in a host population and in which there may be multiple origins of infection, or that can enumerate unobserved infections missing from the sample. Our proposed framework addresses these shortcomings, enabling reconstruction of partially observed transmission trees and estimating the number of cases missing from the sample. Analyses of simulated datasets show the method to be accurate in identifying direct transmissions, while introductions and transmissions via one or more unsampled intermediate cases could be identified at high to moderate levels of case detection. When applied to partial genome sequences of rabies virus sampled from an endemic region of South Africa, our method reveals several distinct transmission cycles with little contact between them, and direct transmission over long distances suggesting significant anthropogenic influence in the movement of infected dogs.