The coevolutionary mosaic of bat betacoronavirus emergence risk.

Pathogen evolution is one of the least predictable components of disease emergence, particularly in nature. Here, building on principles established by the geographic mosaic theory of coevolution, we develop a quantitative, spatially explicit framework for mapping the evolutionary risk of viral emergence. Driven by interest in diseases like Severe Acute Respiratory Syndrome (SARS), Middle East Respiratory Syndrome (MERS), and Coronavirus disease 2019 (COVID-19), we examine the global biogeography of bat-origin betacoronaviruses, and find that coevolutionary principles suggest geographies of risk that are distinct from the hotspots and coldspots of host richness. Further, our framework helps explain patterns like a unique pool of merbecoviruses in the Neotropics, a recently discovered lineage of divergent nobecoviruses in Madagascar, and-most importantly-hotspots of diversification in southeast Asia, sub-Saharan Africa, and the Middle East that correspond to the site of previous zoonotic emergence events. Our framework may help identify hotspots of future risk that have also been previously overlooked, like West Africa and the Indian subcontinent, and may more broadly help researchers understand how host ecology shapes the evolution and diversity of pandemic threats.

Postdoctoral scientists are mentors, and it is time to recognize their work.

Academia often fails to recognize the important work that supports its functioning, such as mentoring and teaching performed by postdoctoral researchers. This is a particular problem for early-career researchers, but opportunities exist to improve the status quo.

What constrains food webs? A maximum entropy framework for predicting their structure with minimal biases.

Food webs are complex ecological networks whose structure is both ecologically and statistically constrained, with many network properties being correlated with each other. Despite the recognition of these invariable relationships in food webs, the use of the principle of maximum entropy (MaxEnt) in network ecology is still rare. This is surprising considering that MaxEnt is a statistical tool precisely designed for understanding and predicting many types of constrained systems. This principle asserts that the least-biased probability distribution of a system's property, constrained by prior knowledge about that system, is the one with maximum information entropy. MaxEnt has been proven useful in many ecological modeling problems, but its application in food webs and other ecological networks is limited. Here we show how MaxEnt can be used to derive many food-web properties both analytically and heuristically. First, we show how the joint degree distribution (the joint probability distribution of the numbers of prey and predators for each species in the network) can be derived analytically using the number of species and the number of interactions in food webs. Second, we present a heuristic and flexible approach of finding a network's adjacency matrix (the network's representation in matrix format) based on simulated annealing and SVD entropy. We built two heuristic models using the connectance and the joint degree sequence as statistical constraints, respectively. We compared both models' predictions against corresponding null and neutral models commonly used in network ecology using open access data of terrestrial and aquatic food webs sampled globally (N = 257). We found that the heuristic model constrained by the joint degree sequence was a good predictor of many measures of food-web structure, especially the nestedness and motifs distribution. Specifically, our results suggest that the structure of terrestrial and aquatic food webs is mainly driven by their joint degree distribution.

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.

Shortcomings of reusing species interaction networks created by different sets of researchers.

Given the requisite cost associated with observing species interactions, ecologists often reuse species interaction networks created by different sets of researchers to test their hypotheses regarding how ecological processes drive network topology. Yet, topological properties identified across these networks may not be sufficiently attributable to ecological processes alone as often assumed. Instead, much of the totality of topological differences between networks-topological heterogeneity-could be due to variations in research designs and approaches that different researchers use to create each species interaction network. To evaluate the degree to which this topological heterogeneity is present in available ecological networks, we first compared the amount of topological heterogeneity across 723 species interaction networks created by different sets of researchers with the amount quantified from non-ecological networks known to be constructed following more consistent approaches. Then, to further test whether the topological heterogeneity was due to differences in study designs, and not only to inherent variation within ecological networks, we compared the amount of topological heterogeneity between species interaction networks created by the same sets of researchers (i.e., networks from the same publication) with the amount quantified between networks that were each from a unique publication source. We found that species interaction networks are highly topologically heterogeneous: while species interaction networks from the same publication are much more topologically similar to each other than interaction networks that are from a unique publication, they still show at least twice as much heterogeneity as any category of non-ecological networks that we tested. Altogether, our findings suggest that extra care is necessary to effectively analyze species interaction networks created by different researchers, perhaps by controlling for the publication source of each network.

Mismatch between IUCN range maps and species interactions data illustrated using the Serengeti food web.

Background: Range maps are a useful tool to describe the spatial distribution of species. However, they need to be used with caution, as they essentially represent a rough approximation of a species' suitable habitats. When stacked together, the resulting communities in each grid cell may not always be realistic, especially when species interactions are taken into account. Here we show the extent of the mismatch between range maps, provided by the International Union for Conservation of Nature (IUCN), and species interactions data. More precisely, we show that local networks built from those stacked range maps often yield unrealistic communities, where species of higher trophic levels are completely disconnected from primary producers. Methodology: We used the well-described Serengeti food web of mammals and plants as our case study, and identify areas of data mismatch within predators' range maps by taking into account food web structure. We then used occurrence data from the Global Biodiversity Information Facility (GBIF) to investigate where data is most lacking. Results: We found that most predator ranges comprised large areas without any overlapping distribution of their prey. However, many of these areas contained GBIF occurrences of the predator. Conclusions: Our results suggest that the mismatch between both data sources could be due either to the lack of information about ecological interactions or the geographical occurrence of prey. We finally discuss general guidelines to help identify defective data among distributions and interactions data, and we recommend this method as a valuable way to assess whether the occurrence data that are being used, even if incomplete, are ecologically accurate.

The Global Virome in One Network (VIRION): an Atlas of Vertebrate-Virus Associations.

Data that catalogue viral diversity on Earth have been fragmented across sources, disciplines, formats, and various degrees of open sharing, posing challenges for research on macroecology, evolution, and public health. Here, we solve this problem by establishing a dynamically maintained database of vertebrate-virus associations, called The Global Virome in One Network (VIRION). The VIRION database has been assembled through both reconciliation of static data sets and integration of dynamically updated databases. These data sources are all harmonized against one taxonomic backbone, including metadata on host and virus taxonomic validity and higher classification; additional metadata on sampling methodology and evidence strength are also available in a harmonized format. In total, the VIRION database is the largest open-source, open-access database of its kind, with roughly half a million unique records that include 9,521 resolved virus "species" (of which 1,661 are ICTV ratified), 3,692 resolved vertebrate host species, and 23,147 unique interactions between taxonomically valid organisms. Together, these data cover roughly a quarter of mammal diversity, a 10th of bird diversity, and ∼6% of the estimated total diversity of vertebrates, and a much larger proportion of their virome than any previous database. We show how these data can be used to test hypotheses about microbiology, ecology, and evolution and make suggestions for best practices that address the unique mix of evidence that coexists in these data. IMPORTANCE Animals and their viruses are connected by a sprawling, tangled network of species interactions. Data on the host-virus network are available from several sources, which use different naming conventions and often report metadata in different levels of detail. VIRION is a new database that combines several of these existing data sources, reconciles taxonomy to a single consistent backbone, and reports metadata in a format designed by and for virologists. Researchers can use VIRION to easily answer questions like "Can any fish viruses infect humans?" or "Which bats host coronaviruses?" or to build more advanced predictive models, making it an unprecedented step toward a full inventory of the global virome.

Optimising predictive models to prioritise viral discovery in zoonotic reservoirs.

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.

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.

Refocusing multiple stressor research around the targets and scales of ecological impacts.

Ecological communities face a variety of environmental and anthropogenic stressors acting simultaneously. Stressor impacts can combine additively or can interact, causing synergistic or antagonistic effects. Our knowledge of when and how interactions arise is limited, as most models and experiments only consider the effect of a small number of non-interacting stressors at one or few scales of ecological organization. This is concerning because it could lead to significant underestimations or overestimations of threats to biodiversity. Furthermore, stressors have been largely classified by their source rather than by the mechanisms and ecological scales at which they act (the target). Here, we argue, first, that a more nuanced classification of stressors by target and ecological scale can generate valuable new insights and hypotheses about stressor interactions. Second, that the predictability of multiple stressor effects, and consistent patterns in their impacts, can be evaluated by examining the distribution of stressor effects across targets and ecological scales. Third, that a variety of existing mechanistic and statistical modelling tools can play an important role in our framework and advance multiple stressor research.

A roadmap towards predicting species interaction networks (across space and time).

Networks of species interactions underpin numerous ecosystem processes, but comprehensively sampling these interactions is difficult. Interactions intrinsically vary across space and time, and given the number of species that compose ecological communities, it can be tough to distinguish between a true negative (where two species never interact) from a false negative (where two species have not been observed interacting even though they actually do). Assessing the likelihood of interactions between species is an imperative for several fields of ecology. This means that to predict interactions between species-and to describe the structure, variation, and change of the ecological networks they form-we need to rely on modelling tools. Here, we provide a proof-of-concept, where we show how a simple neural network model makes accurate predictions about species interactions given limited data. We then assess the challenges and opportunities associated with improving interaction predictions, and provide a conceptual roadmap forward towards predictive models of ecological networks that is explicitly spatial and temporal. We conclude with a brief primer on the relevant methods and tools needed to start building these models, which we hope will guide this research programme forward. This article is part of the theme issue 'Infectious disease macroecology: parasite diversity and dynamics across the globe'.

Optimal transportation theory for species interaction networks.

Observed biotic interactions between species, such as in pollination, predation, and competition, are determined by combinations of population densities, matching in functional traits and phenology among the organisms, and stochastic events (neutral effects).We propose optimal transportation theory as a unified view for modeling species interaction networks with different intensities of interactions. We pose the coupling of two distributions as a constrained optimization problem, maximizing both the system's average utility and its global entropy, that is, randomness. Our model follows naturally from applying the MaxEnt principle to this problem setting.This approach allows for simulating changes in species relative densities as well as to disentangle the impact of trait matching and neutral forces.We provide a framework for estimating the pairwise species utilities from data. Experimentally, we show how to use this framework to perform trait matching and predict the coupling in pollination and host-parasite networks.

Beta and phylogenetic diversities tell complementary stories about ecological networks biogeography.

The beta-diversity of interactions between communities does not necessarily correspond to the differences related to their species composition because interactions show greater variability than species co-occurrence. Additionally, the structure of species interaction networks can itself vary over spatial gradients, thereby adding constraints on the dissimilarity of communities in space. We used published data on the parasitism interaction between fleas and small mammals in 51 regions of the Palearctic to investigate how beta-diversity of networks and phylogenetic diversity are related. The networks could be separated in groups based on the metrics that best described the differences between them, and these groups were also geographically structured. We also found that each network beta-diversity index relates in a particular way with phylogenetically community dissimilarity, reinforcing that some of these indexes have a strong phylogenetic component. Our results clarify important aspects of the biogeography of hosts and parasites communities in Eurasia, while suggesting that networks beta-diversity and phylogenetic dissimilarity interact with the environment in different ways.

Environment-host-microbial interactions shape the Sarracenia purpurea microbiome at the continental scale.

The importance of climate, habitat structure, and higher trophic levels on microbial diversity is only beginning to be understood. Here, we examined the influence of climate variables, plant morphology, and the abundance of aquatic invertebrates on the microbial biodiversity of the northern pitcher plant Sarracenia purpurea. The plant's cup-shaped leaves fill with rainwater and support a miniature, yet full-fledged, ecosystem with a diverse microbiome that decomposes captured prey and a small network of shredding and filter-feeding aquatic invertebrates that feed on microbes. We characterized pitcher microbiomes of 108 plants sampled at 36 sites from Florida to Quebec. Structural equation models revealed that annual precipitation and temperature, plant size, and midge abundance had direct effects on microbiome taxonomic and phylogenetic diversity. Climate variables also exerted indirect effects through plant size and midge abundance. Further, spatial structure and climate influenced taxonomic composition, but not phylogenetic composition. Our results suggest that direct effects of midge abundance and climate and indirect effects of climate through its effect on plant-associated factors lead to greater richness of microbial phylotypes in warmer, wetter sites.

Revisiting the Links-Species Scaling Relationship in Food Webs.

Predicting the number of interactions among species in a food web is an important task. These trophic interactions underlie many ecological and evolutionary processes, ranging from biomass fluxes, ecosystem stability, resilience to extinction, and resistance against novel species. We investigate and compare several ways to predict the number of interactions in food webs. We conclude that a simple beta-binomial model outperforms other models, with the added desirable property of respecting biological constraints. We show how this simple relationship gives rise to a predicted distribution of several quantities related to link number in food webs, including the scaling of network structure with space and the probability that a network will be stable.

Testing predictability of disease outbreaks with a simple model of pathogen biogeography.

Predicting disease emergence and outbreak events is a critical task for public health professionals and epidemiologists. Advances in global disease surveillance are increasingly generating datasets that are worth more than their component parts for prediction-oriented work. Here, we use a trait-free approach which leverages information on the global community of human infectious diseases to predict the biogeography of pathogens through time. Our approach takes pairwise dissimilarities between countries' pathogen communities and pathogens' geographical distributions and uses these to predict country-pathogen associations. We compare the success rates of our model for predicting pathogen outbreak, emergence and re-emergence potential as a function of time (e.g. number of years between training and prediction), pathogen type (e.g. virus) and transmission mode (e.g. vector-borne). With only these simple predictors, our model successfully predicts basic network structure up to a decade into the future. We find that while outbreak and re-emergence potential are especially well captured by our simple model, prediction of emergence events remains more elusive, and sudden global emergences like an influenza pandemic are beyond the predictive capacity of the model. However, these stochastic pandemic events are unlikely to be predictable from such coarse data. Together, our model is able to use the information on the existing country-pathogen network to predict pathogen outbreaks fairly well, suggesting the importance in considering information on co-occurring pathogens in a more global view even to estimate outbreak events in a single location or for a single pathogen.

Revealing biases in the sampling of ecological interaction networks.

The structure of ecological interactions is commonly understood through analyses of interaction networks. However, these analyses may be sensitive to sampling biases with respect to both the interactors (the nodes of the network) and interactions (the links between nodes), because the detectability of species and their interactions is highly heterogeneous. These ecological and statistical issues directly affect ecologists' abilities to accurately construct ecological networks. However, statistical biases introduced by sampling are difficult to quantify in the absence of full knowledge of the underlying ecological network's structure. To explore properties of large-scale ecological networks, we developed the software EcoNetGen, which constructs and samples networks with predetermined topologies. These networks may represent a wide variety of communities that vary in size and types of ecological interactions. We sampled these networks with different mathematical sampling designs that correspond to methods used in field observations. The observed networks generated by each sampling process were then analyzed with respect to the number of components, size of components and other network metrics. We show that the sampling effort needed to estimate underlying network properties depends strongly both on the sampling design and on the underlying network topology. In particular, networks with random or scale-free modules require more complete sampling to reveal their structure, compared to networks whose modules are nested or bipartite. Overall, modules with nested structure were the easiest to detect, regardless of the sampling design used. Sampling a network starting with any species that had a high degree (e.g., abundant generalist species) was consistently found to be the most accurate strategy to estimate network structure. Because high-degree species tend to be generalists, abundant in natural communities relative to specialists, and connected to each other, sampling by degree may therefore be common but unintentional in empirical sampling of networks. Conversely, sampling according to module (representing different interaction types or taxa) results in a rather complete view of certain modules, but fails to provide a complete picture of the underlying network. To reduce biases introduced by sampling methods, we recommend that these findings be incorporated into field design considerations for projects aiming to characterize large species interaction networks.

The marine fish food web is globally connected.

The productivity of marine ecosystems and the services they provide to humans are largely dependent on complex interactions between prey and predators. These are embedded in a diverse network of trophic interactions, resulting in a cascade of events following perturbations such as species extinction. The sheer scale of oceans, however, precludes the characterization of marine feeding networks through de novo sampling. This effort ought instead to rely on a combination of extensive data and inference. Here we investigate how the distribution of trophic interactions at the global scale shapes the marine fish food web structure. We hypothesize that the heterogeneous distribution of species ranges in biogeographic regions should concentrate interactions in the warmest areas and within species groups. We find that the inferred global metaweb of marine fish-that is, all possible potential feeding links between co-occurring species-is highly connected geographically with a low degree of spatial modularity. Metrics of network structure correlate with sea surface temperature and tend to peak towards the tropics. In contrast to open-water communities, coastal food webs have greater interaction redundancy, which may confer robustness to species extinction. Our results suggest that marine ecosystems are connected yet display some resistance to perturbations because of high robustness at most locations.

Ecological Data Should Not Be So Hard to Find and Reuse.

Drawing upon the data deposited in publicly shared archives has the potential to transform the way we conduct ecological research. For this transformation to happen, we argue that data need to be more interoperable and easier to discover. One way to achieve these goals is to adopt domain-specific data representations.