Identifying keystone species in the human gut microbiome from metagenomic timeseries using sparse linear regression

Estimating diversity in networked ecological communities

Comparing ecological communities across environmental gradients can be challenging, especially when the number of different taxonomic groups in the communities is large. In this setting, community-level summaries called diversity indices are widely used to detect changes in the community ecology. However, estimation of diversity indices has received relatively little attention from the statistical community. The most common estimates of diversity are the maximum likelihood estimates of the parameters of a multinomial model, even though the multinomial model implies strict assumptions about the sampling mechanism. In particular, the multinomial model prohibits ecological networks, where taxa positively and negatively co-occur. In this article, we leverage models from the compositional data literature that explicitly account for co-occurrence networks and use them to estimate diversity. Instead of proposing new diversity indices, we estimate popular diversity indices under these models. While the methodology is general, we illustrate the approach for the estimation of the Shannon, Simpson, Bray–Curtis, and Euclidean diversity indices. We contrast our method to multinomial, low-rank, and nonparametric methods for estimating diversity indices. Under simulation, we find that the greatest gains of the method are in strongly networked communities with many taxa. Therefore, to illustrate the method, we analyze the microbiome of seafloor basalts based on a 16S amplicon sequencing dataset with 1425 taxa and 12 communities.

Reconstructing ecological networks with noisy dynamics

Ecosystems functioning is based on an intricate web of interactions among living entities. Most of these interactions are difficult to observe, especially when the diversity of interacting entities is large and they are of small size and abundance. To sidestep this limitation, it has become common to infer the network structure of ecosystems from time series of species abundance, but it is not clear how well can networks be reconstructed, especially in the presence of stochasticity that propagates through ecological networks. We evaluate the effects of intrinsic noise and network topology on the performance of different methods of inferring network structure from time-series data. Analysis of seven different four-species motifs using a stochastic model demonstrates that star-shaped motifs are differentially detected by these methods while rings are differentially constructed. The ability to reconstruct the network is unaffected by the magnitude of stochasticity in the population dynamics. Instead, interaction between the stochastic and deterministic parts of the system determines the path that the whole system takes to equilibrium and shapes the species covariance. We highlight the effects of long transients on the path to equilibrium and suggest a path forward for developing more ecologically sound statistical techniques.

A Primer for Microbiome Time-Series Analysis

Time-series can provide critical insights into the structure and function of microbial communities. The analysis of temporal data warrants statistical considerations, distinct from comparative microbiome studies, to address ecological questions. This primer identifies unique challenges and approaches for analyzing microbiome time-series. In doing so, we focus on (1) identifying compositionally similar samples, (2) inferring putative interactions among populations, and (3) detecting periodic signals. We connect theory, code and data via a series of hands-on modules with a motivating biological question centered on marine microbial ecology. The topics of the modules include characterizing shifts in community structure and activity, identifying expression levels with a diel periodic signal, and identifying putative interactions within a complex community. Modules are presented as self-contained, open-access, interactive tutorials in R and Matlab. Throughout, we highlight statistical considerations for dealing with autocorrelated and compositional data, with an eye to improving the robustness of inferences from microbiome time-series. In doing so, we hope that this primer helps to broaden the use of time-series analytic methods within the microbial ecology research community.

Model-based Comparisons of the Abundance Dynamics of Bacterial Communities in Two Lakes

Lake Lanier (Georgia, USA) is home to more than 11,000 microbial Operational Taxonomic Units (OTUs), many of which exhibit clear annual abundance patterns. To assess the dynamics of this microbial community, we collected time series data of 16S and 18S rRNA gene sequences, recovered from 29 planktonic shotgun metagenomic datasets. Based on these data, we constructed a dynamic mathematical model of bacterial interactions in the lake and used it to analyze changes in the abundances of OTUs. The model accounts for interactions among 14 sub-communities (SCs), which are composed of OTUs blooming at the same time of the year, and three environmental factors. It captures the seasonal variations in abundances of the SCs quite well. Simulation results suggest that changes in water temperature affect the various SCs differentially and that the timing of perturbations is critical. We compared the model results with published results from Lake Mendota (Wisconsin, USA). These comparative analyses between lakes in two very different geographical locations revealed substantially more cooperation and less competition among species in the warmer Lake Lanier than in Lake Mendota.

Higher-Order Interaction between Species Inhibits Bacterial Invasion of a Phototroph-Predator Microbial Community

The composition of an ecosystem is thought to be important for determining its resistance to invasion. Studies of natural ecosystems, from plant to microbial communities, have found that more diverse communities are more resistant to invasion. In some cases, more diverse communities resist invasion by more completely consuming the resources necessary for the invader. We show that Escherichia coli can successfully invade cultures of the alga Chlamydomonas reinhardtii (phototroph) or the ciliate Tetrahymena thermophila (predator) but cannot invade a community where both are present. The invasion resistance of the algae-ciliate community arises from a higher-order interaction between species (interaction modification) that is unrelated to resource consumption. We show that the mode of this interaction is the algal inhibition of bacterial aggregation, which leaves bacteria vulnerable to predation. This mode requires both the algae and the ciliate to be present and provides an example of invasion resistance through an interaction modification.

Untangling the dynamics of persistence and colonization in microbial communities

A central goal of community ecology is to infer biotic interactions from observed distributions of co-occurring species. Evidence for biotic interactions, however, can be obscured by shared environmental requirements, posing a challenge for statistical inference. Here, we introduce a dynamic statistical model, based on probit regression, that quantifies the effects of spatial and temporal covariance in longitudinal co-occurrence data. We separate the fixed pairwise effects of species occurrences on persistence and colonization rates, a potential signal of direct interactions, from latent pairwise correlations in occurrence, a potential signal of shared environmental responses. We first validate our modeling framework with several simulation studies. Then, we apply the approach to a pressing epidemiological question by examining how human papillomavirus (HPV) types coexist. Our results suggest that while HPV types respond similarly to common host traits, direct interactions are sparse and weak, so that HPV type diversity depends largely on shared environmental drivers. Our modeling approach is widely applicable to microbial communities and provides valuable insights that should lead to more directed hypothesis testing and mechanistic modeling.

Metabolite-mediated modelling of microbial community dynamics captures emergent behaviour more effectively than species-species modelling

Personalized models of the gut microbiome are valuable for disease prevention and treatment. For this, one requires a mathematical model that predicts microbial community composition and the emergent behaviour of microbial communities. We seek a modelling strategy that can capture emergent behaviour when built from sets of universal individual interactions. Our investigation reveals that species-metabolite interaction (SMI) modelling is better able to capture emergent behaviour in community composition dynamics than direct species-species modelling. Using publicly available data, we examine the ability of species-species models and species-metabolite models to predict trio growth experiments from the outcomes of pair growth experiments. We compare quadratic species-species interaction models and quadratic SMI models and conclude that only species-metabolite models have the necessary complexity to explain a wide variety of interdependent growth outcomes. We also show that general species-species interaction models cannot match the patterns observed in community growth dynamics, whereas species-metabolite models can. We conclude that species-metabolite modelling will be important in the development of accurate, clinically useful models of microbial communities.

Interrogation of Internal Workings in Microbial Community Assembly: Play a Game through a Behavioral Network?

Identifying general biological rules that underlie the complexity and heterogeneity of microbial communities has proven to be highly challenging. We present a rule-of-thumb framework for studying and characterizing how microbes interact with each other across different taxa to determine community behavior and dynamics. This framework is computationally simple but conceptually meaningful, and it can provide a starting point to generate novel biological hypotheses about microbial interactions and explore internal workings of microbial community assembly in depth.

Increasing evidence shows that the influence of microbiota on biogeochemical cycling, plant development, and human health is executed through a complex network of microbe-microbe interactions. However, characterizing how microbes interact and work together within closely packed and highly heterogeneous microbial ecosystems is extremely challenging. Here, we describe a rule-of-thumb framework for visualizing polymicrobial interactions and extracting general principles that underlie microbial communities. We integrate elements of metabolic ecology, behavioral ecology, and game theory to quantify the interactive strategies by which microbes at any taxonomic level compete for resources and cooperate symbiotically with each other to form and stabilize ecological communities. We show how the framework can chart an omnidirectional landscape of microbial cooperation and competition that may drive various natural processes. This framework can be implemented into genome-wide association studies to unravel the genetic mechanisms underlying microbial interaction networks and their evolutionary consequences along spatiotemporal gradients.

IMPORTANCE Identifying general biological rules that underlie the complexity and heterogeneity of microbial communities has proven to be highly challenging. We present a rule-of-thumb framework for studying and characterizing how microbes interact with each other across different taxa to determine community behavior and dynamics. This framework is computationally simple but conceptually meaningful, and it can provide a starting point to generate novel biological hypotheses about microbial interactions and explore internal workings of microbial community assembly in depth.

Detection of multi-dimensional co-exclusion patterns in microbial communities

Motivation

Identification of complex relationships among members of microbial communities is key to understand and control the microbiota. Co-exclusion is arguably one of the most important patterns reflecting micro-organisms' intolerance to each other's presence. Knowing these relations opens an opportunity to manipulate microbiotas, personalize anti-microbial and probiotic treatments as well as guide microbiota transplantation. The co-exclusion pattern however, cannot be appropriately described by a linear function nor its strength be estimated using covariance or (negative) Pearson and Spearman correlation coefficients. This manuscript proposes a way to quantify the strength and evaluate the statistical significance of co-exclusion patterns between two, three or more variables describing a microbiota and allows one to extend analysis beyond micro-organism abundance by including other microbiome associated measurements such as, pH, temperature etc., as well as estimate the expected numbers of false positive co-exclusion patterns in a co-exclusion network.

Results

The implemented computational pipeline (CoEx) tested against 2380 microbial profiles (samples) from The Human Microbiome Project resulted in body-site specific pairwise co-exclusion patterns.

Availability and implementation

C++ source code for calculation of the score and P-value for two, three and four dimensional co-exclusion patterns as well as source code and executable files for the CoEx pipeline are available at https://scsb.utmb.edu/labgroups/fofanov/co-exclusion_in_microbial_communities.asp.

Supplementary information

Supplementary data are available at Bioinformatics online.

Primate microbiomes over time: Longitudinal answers to standing questions in microbiome research

To date, most insights into the processes shaping vertebrate gut microbiomes have emerged from studies with cross-sectional designs. While this approach has been valuable, emerging time series analyses on vertebrate gut microbiomes show that gut microbial composition can change rapidly from 1 day to the next, with consequences for host physical functioning, health, and fitness. Hence, the next frontier of microbiome research will require longitudinal perspectives. Here we argue that primatologists, with their traditional focus on tracking the lives of individual animals and familiarity with longitudinal fecal sampling, are well positioned to conduct research at the forefront of gut microbiome dynamics. We begin by reviewing some of the most important ecological processes governing microbiome change over time, and briefly summarizing statistical challenges and approaches to microbiome time series analysis. We then introduce five questions of general interest to microbiome science where we think field-based primate studies are especially well positioned to fill major gaps: (a) Do early life events shape gut microbiome composition in adulthood? (b) Do shifting social landscapes cause gut microbial change? (c) Are gut microbiome phenotypes heritable across variable environments? (d) Does the gut microbiome show signs of host aging? And (e) do gut microbiome composition and dynamics predict host health and fitness? For all of these questions, we highlight areas where primatologists are uniquely positioned to make substantial contributions. We review preliminary evidence, discuss possible study designs, and suggest future directions.

Microbial Interaction Network Inference in Microfluidic Droplets

Microbial interactions are major drivers of microbial community dynamics and functions but remain challenging to identify because of limitations in parallel culturing and absolute abundance quantification of community members across environments and replicates. To this end, we developed Microbial Interaction Network Inference in microdroplets (MINI-Drop). Fluorescence microscopy coupled to computer vision techniques were used to rapidly determine the absolute abundance of each strain in hundreds to thousands of droplets per condition. We showed that MINI-Drop could accurately infer pairwise and higher-order interactions in synthetic consortia. We developed a stochastic model of community assembly to provide insight into the heterogeneity in community states across droplets. Finally, we elucidated the complex web of interactions linking antibiotics and different species in a synthetic consortium. In sum, we demonstrated a robust and generalizable method to infer microbial interaction networks by random encapsulation of sub-communities into microfluidic droplets.

Consumer Safety Considerations of Skin and Oral Microbiome Perturbation

Microbiomes associated with human skin and the oral cavity are uniquely exposed to personal care regimes. Changes in the composition and activities of the microbial communities in these environments can be utilized to promote consumer health benefits, for example, by reducing the numbers, composition, or activities of microbes implicated in conditions such as acne, axillary odor, dandruff, and oral diseases. It is, however, important to ensure that innovative approaches for microbiome manipulation do not unsafely disrupt the microbiome or compromise health, and where major changes in the composition or activities of the microbiome may occur, these require evaluation to ensure that critical biological functions are unaffected. This article is based on a 2-day workshop held at SEAC Unilever, Sharnbrook, United Kingdom, involving 31 specialists in microbial risk assessment, skin and oral microbiome research, microbial ecology, bioinformatics, mathematical modeling, and immunology. The first day focused on understanding the potential implications of skin and oral microbiome perturbation, while approaches to characterize those perturbations were discussed during the second day. This article discusses the factors that the panel recommends be considered for personal care products that target the microbiomes of the skin and the oral cavity.

An expectation-maximization algorithm enables accurate ecological modeling using longitudinal microbiome sequencing data

BACKGROUND

The dynamics of microbial communities is driven by a range of interactions from symbiosis to predator-prey relationships, the majority of which are poorly understood. With the increasing availability of high-throughput microbiome taxonomic profiling data, it is now conceivable to directly learn the ecological models that explicitly define microbial interactions and explain community dynamics. The applicability of these approaches is severely limited by the lack of accurate absolute cell density measurements (biomass).

METHODS

We present a new computational approach that resolves this key limitation in the inference of generalized Lotka-Volterra models (gLVMs) by coupling biomass estimation and model inference with an expectation-maximization algorithm (BEEM).

RESULTS

BEEM outperforms the state-of-the-art methods for inferring gLVMs, while simultaneously eliminating the need for additional experimental biomass data as input. BEEM's application to previously inaccessible public datasets (due to the lack of biomass data) allowed us to construct ecological models of microbial communities in the human gut on a per-individual basis, revealing personalized dynamics and keystone species.

CONCLUSIONS

BEEM addresses a key bottleneck in "systems analysis" of microbiomes by enabling accurate inference of ecological models from high throughput sequencing data without the need for experimental biomass measurements.

Timescales of gut microbiome dynamics

Vast communities of microorganisms inhabit the gastrointestinal tracts of humans and other animals. Understanding their initial development, fluctuations in composition, stability over long times, and responses to transient perturbations - in other words their dynamics - is important both for gaining basic insights into these ecosystems and for rationally manipulating them for therapeutic ends. Gut microbiome dynamics, however, remain poorly understood. We review here studies of gut microbiome dynamics in the presence and absence of external perturbations, noting especially the long timescales associated with overall stability and the short timescales associated with various underlying biological processes. Integrating these disparate timescales, we suggest, is an important goal for future work and is necessary for developing a predictive understanding of microbiome dynamics.

The Dynamic of a River Model Bacterial Community in Two Different Media Reveals a Divergent Succession and an Enhanced Growth of Most Strains Compared to Monocultures

The dynamic of a community of 20 bacterial strains isolated from river water was followed in R2 broth and in autoclaved river water medium for 27 days in batch experiments. At an early stage of incubation, a fast-growing specialist strain, Acinetobater sp., dominated the community in both media. Later on, the community composition in both media diverged but was highly reproducible across replicates. In R2, several strains previously reported to degrade multiple simple carbon sources prevailed. In autoclaved river water, the community was more even and became dominated by several strains growing faster or exclusively in that medium. Those strains have been reported in the literature to degrade complex compounds. Their growth rate in the community was 1.5- to 7-fold greater than that observed in monoculture. Furthermore, those strains developed simultaneously in the community. Together, our results suggest the existence of cooperative interactions within the community incubated in autoclaved river water.

Use and abuse of correlation analyses in microbial ecology

Correlation analyses are often included in bioinformatic pipelines as methods for inferring taxon-taxon interactions. In this perspective, we highlight the pitfalls of inferring interactions from covariance and suggest methods, study design considerations, and additional data types for improving high-throughput interaction inferences. We conclude that correlation, even when augmented by other data types, almost never provides reliable information on direct biotic interactions in real-world ecosystems. These bioinformatically inferred associations are useful for reducing the number of potential hypotheses that we might test, but will never preclude the necessity for experimental validation.

Detecting interaction networks in the human microbiome with conditional Granger causality

Human microbiome research is rife with studies attempting to deduce microbial correlation networks from sequencing data. Standard correlation and/or network analyses may be misleading when taken as an indication of taxon interactions because "correlation is neither necessary nor sufficient to establish causation"; environmental filtering can lead to correlation between non-interacting taxa. Unfortunately, microbial ecologists have generally used correlation as a proxy for causality although there is a general consensus about what constitutes a causal relationship: causes both precede and predict effects. We apply one of the first causal models for detecting interactions in human microbiome samples. Specifically, we analyze a long duration, high resolution time series of the human microbiome to decipher the networks of correlation and causation of human-associated microbial genera. We show that correlation is not a good proxy for biological interaction; we observed a weak negative relationship between correlation and causality. Strong interspecific interactions are disproportionately positive, whereas almost all strong intraspecific interactions are negative. Interestingly, intraspecific interactions also appear to act at a short timescale causing vast majority of the effects within 1-3 days. We report how different taxa are involved in causal relationships with others, and show that strong interspecific interactions are rarely conserved across two body sites whereas strong intraspecific interactions are much more conserved, ranging from 33% between the gut and right-hand to 70% between the two hands. Therefore, in the absence of guiding assumptions about ecological interactions, Granger causality and related techniques may be particularly helpful for understanding the driving factors governing microbiome composition and structure.

Mapping the microbial interactome: Statistical and experimental approaches for microbiome network inference

IMPACT STATEMENT

This review provides a comprehensive description of experimental and statistical tools used for network analyses of the human gut microbiome. Understanding the system dynamics of microbial interactions may lead to the improvement of therapeutic approaches for managing microbiome-associated diseases. Microbiome network inference tools have been developed and applied to both cross-sectional and longitudinal experimental designs, as well as to multi-omic datasets, with the goal of untangling the complex web of microbe-host, microbe-environmental, and metabolism-mediated microbial interactions. The characterization of these interaction networks may lead to a better understanding of the systems dynamics of the human gut microbiome, augmenting our knowledge of the microbiome's role in human health, and guiding the optimization of effective, precise, and rational therapeutic strategies for managing microbiome-associated disease.

Web-gLV: A Web Based Platform for Lotka-Volterra Based Modeling and Simulation of Microbial Populations

The affordability of high throughput DNA sequencing has allowed us to explore the dynamics of microbial populations in various ecosystems. Mathematical modeling and simulation of such microbiome time series data can help in getting better understanding of bacterial communities. In this paper, we present Web-gLV-a GUI based interactive platform for generalized Lotka-Volterra (gLV) based modeling and simulation of microbial populations. The tool can be used to generate the mathematical models with automatic estimation of parameters and use them to predict future trajectories using numerical simulations. We also demonstrate the utility of our tool on few publicly available datasets. The case studies demonstrate the ease with which the current tool can be used by biologists to model bacterial populations and simulate their dynamics to get biological insights. We expect Web-gLV to be a valuable contribution in the field of ecological modeling and metagenomic systems biology.

From hairballs to hypotheses-biological insights from microbial networks

Microbial networks are an increasingly popular tool to investigate microbial community structure, as they integrate multiple types of information and may represent systems-level behaviour. Interpreting these networks is not straightforward, and the biological implications of network properties are unclear. Analysis of microbial networks allows researchers to predict hub species and species interactions. Additionally, such analyses can help identify alternative community states and niches. Here, we review factors that can result in spurious predictions and address emergent properties that may be meaningful in the context of the microbiome. We also give an overview of studies that analyse microbial networks to identify new hypotheses. Moreover, we show in a simulation how network properties are affected by tool choice and environmental factors. For example, hub species are not consistent across tools, and environmental heterogeneity induces modularity. We highlight the need for robust microbial network inference and suggest strategies to infer networks more reliably.

Mutations in bacterial genes induce unanticipated changes in the relationship between bacterial pathogens in experimental otitis media

Otitis media (OM) is a common polymicrobial infection of the middle ear in children under the age of 15 years. A widely used experimental strategy to analyse roles of specific phenotypes of bacterial pathogens of OM is to study changes in co-infection kinetics of bacterial populations in animal models when a wild-type bacterial strain is replaced by a specific isogenic mutant strain in the co-inoculating mixtures. As relationships between the OM bacterial pathogens within the host are regulated by many interlinked processes, connecting the changes in the co-infection kinetics to a bacterial phenotype can be challenging. We investigated middle ear co-infections in adult chinchillas (Chinchilla lanigera) by two major OM pathogens: non-typeable Haemophilus influenzae (NTHi) and Moraxella catarrhalis (Mcat), as well as isogenic mutant strains in each bacterial species. We analysed the infection kinetic data using Lotka-Volterra population dynamics, maximum entropy inference and Akaike information criteria-(AIC)-based model selection. We found that changes in relationships between the bacterial pathogens that were not anticipated in the design of the co-infection experiments involving mutant strains are common and were strong regulators of the co-infecting bacterial populations. The framework developed here allows for a systematic analysis of host-host variations of bacterial populations and small sizes of animal cohorts in co-infection experiments to quantify the role of specific mutant strains in changing the infection kinetics. Our combined approach can be used to analyse the functional footprint of mutant strains in regulating co-infection kinetics in models of experimental OM and other polymicrobial diseases.

Inference of Significant Microbial Interactions From Longitudinal Metagenomics Data

Data of next-generation sequencing (NGS) and their analysis have been facilitating advances in our understanding of microbial ecosystems such as human gut microbiota. However, inference of microbial interactions occurring within an ecosystem is still a challenge mainly due to sequencing data (e.g., 16S rDNA sequences) providing relative abundance of microbes instead of absolute cell count. In order to describe growtth dynamics of microbial communities and estimate the involved microbial interactions, we introduce a procedure by integrating generalized Lotka-Volterra equations, forward stepwise regression and bootstrap aggregation. First, we successfully identify experimentally confirmed microbial interactions based on relative abundance data of a cheese microbial community. Then, we apply the procedure to time-series of 16S rDNA sequences of gut microbiomes of children who were progressing to Type 1 diabetes (T1D progressors), and compare their gut microbial interactions to a healthy control group. Our results suggest that the number of inferred microbial interactions increased over time during the first 3 years of life. More microbial interactions are found in the gut flora of healthy children than that of T1D progressors. The inhibitory effects from Actinobacteria and Bacilli to Bacteroidia, from Bacteroidia to Clostridia, and the beneficial effect from Clostridia to Bacteroidia are shared between healthy children and T1D progressors. An inhibition of Clostridia by Gammaproteobacteria is found in healthy children that maintains through their first 3 years of life. This suppression appears in T1D progressors during the first year of life, which transforms to a commensalism relationship at the age of 3 years old. Gammaproteobacteria is found exerting an inhibition on Bacteroidia in the T1D progressors, which is not identified in the healthy controls.

One Health Relationships Between Human, Animal, and Environmental Microbiomes: A Mini-Review

The One Health concept stresses the ecological relationships between human, animal, and environmental health. Much of the One Health literature to date has examined the transfer of pathogens from animals (e.g., emerging zoonoses) and the environment to humans. The recent rapid development of technology to perform high throughput DNA sequencing has expanded this view to include the study of entire microbial communities. Applying the One Health approach to the microbiome allows for consideration of both pathogenic and non-pathogenic microbial transfer between humans, animals, and the environment. We review recent research studies of such transmission, the molecular and statistical methods being used, and the implications of such microbiome relationships for human health. Our review identified evidence that the environmental microbiome as well as the microbiome of animals in close contact can affect both the human microbiome and human health outcomes. Such microbiome transfer can take place in the household as well as the workplace setting. Urbanization of built environments leads to changes in the environmental microbiome which could be a factor in human health. While affected by environmental exposures, the human microbiome also can modulate the response to environmental factors through effects on metabolic and immune function. Better understanding of these microbiome interactions between humans, animals, and the shared environment will require continued development of improved statistical and ecological modeling approaches. Such enhanced understanding could lead to innovative interventions to prevent and manage a variety of human health and disease states.

Limitations of Correlation-Based Inference in Complex Virus-Microbe Communities

Inferring interactions from population time series is an active and ongoing area of research. It is relevant across many biological systems—particularly in virus-microbe communities, but also in gene regulatory networks, neural networks, and ecological communities broadly. Correlation-based inference—using correlations to predict interactions—is widespread. However, it is well-known that “correlation does not imply causation.” Despite this, many studies apply correlation-based inference methods to experimental time series without first assessing the potential scope for accurate inference. Here, we find that several correlation-based inference methods fail to recover interactions within in silico virus-microbe communities, raising questions on their relevance when applied in situ.

Microbes are present in high abundances in the environment and in human-associated microbiomes, often exceeding 1 million per ml. Viruses of microbes are present in even higher abundances and are important in shaping microbial populations, communities, and ecosystems. Given the relative specificity of viral infection, it is essential to identify the functional linkages between viruses and their microbial hosts, particularly given dynamic changes in virus and host abundances. Multiple approaches have been proposed to infer infection networks from time series of in situ communities, among which correlation-based approaches have emerged as the de facto standard. In this work, we evaluate the accuracy of correlation-based inference methods using an in silico approach. In doing so, we compare predicted networks to actual networks to assess the self-consistency of correlation-based inference. At odds with assumptions underlying its widespread use, we find that correlation is a poor predictor of interactions in the context of viral infection and lysis of microbial hosts. The failure to predict interactions holds for methods that leverage product-moment, time-lagged, and relative-abundance-based correlations. In closing, we discuss alternative inference methods, particularly model-based methods, as a means to infer interactions in complex microbial communities with viruses.

IMPORTANCE Inferring interactions from population time series is an active and ongoing area of research. It is relevant across many biological systems—particularly in virus-microbe communities, but also in gene regulatory networks, neural networks, and ecological communities broadly. Correlation-based inference—using correlations to predict interactions—is widespread. However, it is well-known that “correlation does not imply causation.” Despite this, many studies apply correlation-based inference methods to experimental time series without first assessing the potential scope for accurate inference. Here, we find that several correlation-based inference methods fail to recover interactions within in silico virus-microbe communities, raising questions on their relevance when applied in situ.

Growth tradeoffs produce complex microbial communities on a single limiting resource

The relationship between the dynamics of a community and its constituent pairwise interactions is a fundamental problem in ecology. Higher-order ecological effects beyond pairwise interactions may be key to complex ecosystems, but mechanisms to produce these effects remain poorly understood. Here we model microbial growth and competition to show that higher-order effects can arise from variation in multiple microbial growth traits, such as lag times and growth rates, on a single limiting resource with no other interactions. These effects produce a range of ecological phenomena: an unlimited number of strains can exhibit multistability and neutral coexistence, potentially with a single keystone strain; strains that coexist in pairs do not coexist all together; and a strain that wins all pairwise competitions can go extinct in a mixed competition. Since variation in multiple growth traits is ubiquitous in microbial populations, our results indicate these higher-order effects may also be widespread, especially in laboratory ecology and evolution experiments.

Higher-order interactions occur when one species mediates the interaction between two others. Here, the authors model microbial growth and competition to show that higher-order interactions can arise from tradeoffs in growth traits, leading to neutral coexistence and other complex dynamics.

Linking microbial co-occurrences to soil ecological processes across a woodland-grassland ecotone

Ecotones between distinct ecosystems have been the focus of many studies as they offer valuable insights into key drivers of community structure and ecological processes that underpin function. While previous studies have examined a wide range of above-ground parameters in ecotones, soil microbial communities have received little attention. Here we investigated spatial patterns, composition, and co-occurrences of archaea, bacteria, and fungi, and their relationships with soil ecological processes across a woodland-grassland ecotone. Geostatistical kriging and network analysis revealed that the community structure and spatial patterns of soil microbiota varied considerably between three habitat components across the ecotone. Woodland samples had significantly higher diversity of archaea while the grassland samples had significantly higher diversity of bacteria. Microbial co-occurrences reflected differences in soil properties and ecological processes. While microbial networks were dominated by bacterial nodes, different ecological processes were linked to specific microbial guilds. For example, soil phosphorus and phosphatase activity formed the largest clusters in their respective networks, and two lignolytic enzymes formed joined clusters. Bacterial ammonia oxidizers were dominant over archaeal oxidizers and showed a significant association (p < 0.001) with potential nitrification (PNR), with the PNR subnetwork being dominated by Betaproteobacteria. The top ten keystone taxa comprised six bacterial and four fungal OTUs, with Random Forest Analysis revealing soil carbon and nitrogen as the determinants of the abundance of keystone taxa. Our results highlight the importance of assessing interkingdom associations in soil microbial networks. Overall, this study shows how ecotones can be used as a model to delineate microbial structural patterns and ecological processes across adjoining land-uses within a landscape.

Microbial communities as dynamical systems

Nowadays, microbial communities are frequently monitored over long periods of time and the interactions between their members are explored in vitro. This development has opened the way to apply mathematical models to characterize community structure and dynamics, to predict responses to perturbations and to explore general dynamical properties such as stability, alternative stable states and periodicity. Here, we highlight the role of dynamical systems theory in the exploration of microbial communities, with a special emphasis on the generalized Lotka-Volterra (gLV) equations. In particular, we discuss applications, assumptions and limitations of the gLV model, mention modifications to address these limitations and review stochastic extensions. The development of dynamical models, together with the generation of time series data, can improve the design and control of microbial communities.

Signatures of ecological processes in microbial community time series

BACKGROUND

Growth rates, interactions between community members, stochasticity, and immigration are important drivers of microbial community dynamics. In sequencing data analysis, such as network construction and community model parameterization, we make implicit assumptions about the nature of these drivers and thereby restrict model outcome. Despite apparent risk of methodological bias, the validity of the assumptions is rarely tested, as comprehensive procedures are lacking. Here, we propose a classification scheme to determine the processes that gave rise to the observed time series and to enable better model selection.

RESULTS

We implemented a three-step classification scheme in R that first determines whether dependence between successive time steps (temporal structure) is present in the time series and then assesses with a recently developed neutrality test whether interactions between species are required for the dynamics. If the first and second tests confirm the presence of temporal structure and interactions, then parameters for interaction models are estimated. To quantify the importance of temporal structure, we compute the noise-type profile of the community, which ranges from black in case of strong dependency to white in the absence of any dependency. We applied this scheme to simulated time series generated with the Dirichlet-multinomial (DM) distribution, Hubbell's neutral model, the generalized Lotka-Volterra model and its discrete variant (the Ricker model), and a self-organized instability model, as well as to human stool microbiota time series. The noise-type profiles for all but DM data clearly indicated distinctive structures. The neutrality test correctly classified all but DM and neutral time series as non-neutral. The procedure reliably identified time series for which interaction inference was suitable. Both tests were required, as we demonstrated that all structured time series, including those generated with the neutral model, achieved a moderate to high goodness of fit to the Ricker model.

CONCLUSIONS

We present a fast and robust scheme to classify community structure and to assess the prevalence of interactions directly from microbial time series data. The procedure not only serves to determine ecological drivers of microbial dynamics, but also to guide selection of appropriate community models for prediction and follow-up analysis.

Deciphering microbial interactions in synthetic human gut microbiome communities

The ecological forces that govern the assembly and stability of the human gut microbiota remain unresolved. We developed a generalizable model-guided framework to predict higher-dimensional consortia from time-resolved measurements of lower-order assemblages. This method was employed to decipher microbial interactions in a diverse human gut microbiome synthetic community. We show that pairwise interactions are major drivers of multi-species community dynamics, as opposed to higher-order interactions. The inferred ecological network exhibits a high proportion of negative and frequent positive interactions. Ecological drivers and responsive recipient species were discovered in the network. Our model demonstrated that a prevalent positive and negative interaction topology enables robust coexistence by implementing a negative feedback loop that balances disparities in monospecies fitness levels. We show that negative interactions could generate history-dependent responses of initial species proportions that frequently do not originate from bistability. Measurements of extracellular metabolites illuminated the metabolic capabilities of monospecies and potential molecular basis of microbial interactions. In sum, these methods defined the ecological roles of major human-associated intestinal species and illuminated design principles of microbial communities.

A Small Molecule-Screening Pipeline to Evaluate the Therapeutic Potential of 2-Aminoimidazole Molecules Against Clostridium difficile

Antibiotics are considered to be the first line of treatment for mild to moderately severe Clostridium difficile infection (CDI) in humans. However, antibiotics are also risk factors for CDI as they decrease colonization resistance against C. difficile by altering the gut microbiota and metabolome. Finding compounds that selectively inhibit different stages of the C. difficile life cycle, while sparing the indigenous gut microbiota is important for the development of alternatives to standard antibiotic treatment. 2-aminoimidazole (2-AI) molecules are known to disrupt bacterial protection mechanisms in antibiotic resistant bacteria such as Pseudomonas aeruginosa, Acinetobacter baumannii, and Staphylococcus aureus, but are yet to be evaluated against C. difficile. A comprehensive small molecule-screening pipeline was developed to investigate how novel small molecules affect different stages of the C. difficile life cycle (growth, toxin, and sporulation) in vitro, and a library of commensal bacteria that are associated with colonization resistance against C. difficile. The initial screening tested the efficacy of eleven 2-AI molecules (compound 1 through 11) against C. difficile R20291 compared to a vancomycin (2 μg/ml) control. Molecules were selected for their ability to inhibit C. difficile growth, toxin activity, and sporulation. Further testing included growth inhibition of other C. difficile strains (CD196, M68, CF5, 630, BI9, M120) belonging to distinct PCR ribotypes, and a commensal panel (Bacteroides fragilis, B. thetaiotaomicron, C. scindens, C. hylemonae, Lactobacillus acidophilus, L. gasseri, Escherichia coli, B. longum subsp. infantis). Three molecules compound 1 and 2, and 3 were microbicidal, whereas compounds 4, 7, 9, and 11 inhibited toxin activity without affecting the growth of C. difficile strains and the commensal microbiota. The antimicrobial and anti-toxin effects of 2-AI molecules need to be further characterized for mode of action and validated in a mouse model of CDI.

Processing and Analyzing Human Microbiome Data

The human microbiome is associated with complex disorders such as diabetes, cancer, obesity and cardiovascular disorders. Recent technological developments have allowed researchers to fully quantify the composition of the microbiome using culture-independent approaches, resulting in a large amount of microbiome data, which provide invaluable opportunities to assess the important contributions of the microbiome to human health and disease. In this chapter, we discuss and evaluate multiple statistical approaches for processing, summarizing, and analyzing microbiome data. Specifically, we provide programming scripts for processing microbiome data using QIIME and calculating alpha and beta diversities, assessing the association between diversities and outcomes of interest using R programs, as well as interpretation of results. We illustrate the methods in the context of analyzing the foregut microbiome in esophageal adenocarcinoma.

Studying microbial functionality within the gut ecosystem by systems biology

Humans are not autonomous entities. We are all living in a complex environment, interacting not only with our peers, but as true holobionts; we are also very much in interaction with our coexisting microbial ecosystems living on and especially within us, in the intestine. Intestinal microorganisms, often collectively referred to as intestinal microbiota, contribute significantly to our daily energy uptake by breaking down complex carbohydrates into simple sugars, which are fermented to short-chain fatty acids and subsequently absorbed by human cells. They also have an impact on our immune system, by suppressing or enhancing the growth of malevolent and beneficial microbes. Our lifestyle can have a large influence on this ecosystem. What and how much we consume can tip the ecological balance in the intestine. A "western diet" containing mainly processed food will have a different effect on our health than a balanced diet fortified with pre- and probiotics. In recent years, new technologies have emerged, which made a more detailed understanding of microbial communities and ecosystems feasible. This includes progress in the sequencing of PCR-amplified phylogenetic marker genes as well as the collective microbial metagenome and metatranscriptome, allowing us to determine with an increasing level of detail, which microbial species are in the microbiota, understand what these microorganisms do and how they respond to changes in lifestyle and diet. These new technologies also include the use of synthetic and in vitro systems, which allow us to study the impact of substrates and addition of specific microbes to microbial communities at a high level of detail, and enable us to gather quantitative data for modelling purposes. Here, we will review the current state of microbiome research, summarizing the computational methodologies in this area and highlighting possible outcomes for personalized nutrition and medicine.

Systems biology of the human microbiome

Recent research has shown that the microbiome-a collection of microorganisms, including bacteria, fungi, and viruses, living on and in a host-are of extraordinary importance in human health, even from conception and development in the uterus. Therefore, to further our ability to diagnose disease, to predict treatment outcomes, and to identify novel therapeutics, it is essential to include microbiome and microbial metabolic biomarkers in Systems Biology investigations. In clinical studies or, more precisely, Systems Medicine approaches, we can use the diversity and individual characteristics of the personal microbiome to enhance our resolution for patient stratification. In this review, we explore several Systems Medicine approaches, including Microbiome Wide Association Studies to understand the role of the human microbiome in health and disease, with a focus on 'preventive medicine' or P4 (i.e., personalized, predictive, preventive, participatory) medicine.

Temporal probabilistic modeling of bacterial compositions derived from 16S rRNA sequencing

Motivation

The number of microbial and metagenomic studies has increased drastically due to advancements in next-generation sequencing-based measurement techniques. Statistical analysis and the validity of conclusions drawn from (time series) 16S rRNA and other metagenomic sequencing data is hampered by the presence of significant amount of noise and missing data (sampling zeros). Accounting uncertainty in microbiome data is often challenging due to the difficulty of obtaining biological replicates. Additionally, the compositional nature of current amplicon and metagenomic data differs from many other biological data types adding another challenge to the data analysis.

Results

To address these challenges in human microbiome research, we introduce a novel probabilistic approach to explicitly model overdispersion and sampling zeros by considering the temporal correlation between nearby time points using Gaussian Processes. The proposed Temporal Gaussian Process Model for Compositional Data Analysis (TGP-CODA) shows superior modeling performance compared to commonly used Dirichlet-multinomial, multinomial and non-parametric regression models on real and synthetic data. We demonstrate that the nonreplicative nature of human gut microbiota studies can be partially overcome by our method with proper experimental design of dense temporal sampling. We also show that different modeling approaches have a strong impact on ecological interpretation of the data, such as stationarity, persistence and environmental noise models.

Availability and implementation

A Stan implementation of the proposed method is available under MIT license at https://github.com/tare/GPMicrobiome.

Contact

taijo@flatironinstitute.org or rb113@nyu.edu.

Supplementary information

Supplementary data are available at Bioinformatics online.

In silico analysis of antibiotic-induced Clostridium difficile infection: Remediation techniques and biological adaptations

In this paper we study antibiotic-induced C. difficile infection (CDI), caused by the toxin-producing C. difficile (CD), and implement clinically-inspired simulated treatments in a computational framework that synthesizes a generalized Lotka-Volterra (gLV) model with SIR modeling techniques. The gLV model uses parameters derived from an experimental mouse model, in which the mice are administered antibiotics and subsequently dosed with CD. We numerically identify which of the experimentally measured initial conditions are vulnerable to CD colonization, then formalize the notion of CD susceptibility analytically. We simulate fecal transplantation, a clinically successful treatment for CDI, and discover that both the transplant timing and transplant donor are relevant to the the efficacy of the treatment, a result which has clinical implications. We incorporate two nongeneric yet dangerous attributes of CD into the gLV model, sporulation and antibiotic-resistant mutation, and for each identify relevant SIR techniques that describe the desired attribute. Finally, we rely on the results of our framework to analyze an experimental study of fecal transplants in mice, and are able to explain observed experimental results, validate our simulated results, and suggest model-motivated experiments.

Oxidative stress in the oral cavity is driven by individual-specific bacterial communities

The term "bacterial dysbiosis" is being used quite extensively in metagenomic studies, however, the identification of harmful bacteria often fails due to large overlap between the bacterial species found in healthy volunteers and patients. We hypothesized that the pathogenic oral bacteria are individual-specific and they correlate with oxidative stress markers in saliva which reflect the inflammatory processes in the oral cavity. Temporally direct and lagged correlations between the markers and bacterial taxa were computed individually for 26 volunteers who provided saliva samples during one month (21.2 ± 2.7 samples/volunteer, 551 samples in total). The volunteers' microbiomes differed significantly by their composition and also by their degree of microbiome temporal variability and oxidative stress markers fluctuation. The results showed that each of the marker-taxa pairs can have negative correlations in some volunteers while positive in others. Streptococcus mutans, which used to be associated with caries before the metagenomics era, had the most prominent correlations with the oxidative stress markers, however, these correlations were not confirmed in all volunteers. The importance of longitudinal samples collections in correlation studies was underlined by simulation of single sample collections in 1000 different combinations which produced contradictory results. In conclusion, the distinct intra-individual correlation patterns suggest that different bacterial consortia might be involved in the oxidative stress induction in each human subject. In the future, decreasing cost of DNA sequencing will allow to analyze multiple samples from each patient, which might help to explore potential diagnostic applications and understand pathogenesis of microbiome-associated oral diseases.

Mapping the ecological networks of microbial communities

Mapping the ecological networks of microbial communities is a necessary step toward understanding their assembly rules and predicting their temporal behavior. However, existing methods require assuming a particular population dynamics model, which is not known a priori. Moreover, those methods require fitting longitudinal abundance data, which are often not informative enough for reliable inference. To overcome these limitations, here we develop a new method based on steady-state abundance data. Our method can infer the network topology and inter-taxa interaction types without assuming any particular population dynamics model. Additionally, when the population dynamics is assumed to follow the classic Generalized Lotka-Volterra model, our method can infer the inter-taxa interaction strengths and intrinsic growth rates. We systematically validate our method using simulated data, and then apply it to four experimental data sets. Our method represents a key step towards reliable modeling of complex, real-world microbial communities, such as the human gut microbiota.

Statistical analysis of co-occurrence patterns in microbial presence-absence datasets

Drawing on a long history in macroecology, correlation analysis of microbiome datasets is becoming a common practice for identifying relationships or shared ecological niches among bacterial taxa. However, many of the statistical issues that plague such analyses in macroscale communities remain unresolved for microbial communities. Here, we discuss problems in the analysis of microbial species correlations based on presence-absence data. We focus on presence-absence data because this information is more readily obtainable from sequencing studies, especially for whole-genome sequencing, where abundance estimation is still in its infancy. First, we show how Pearson's correlation coefficient (r) and Jaccard's index (J)-two of the most common metrics for correlation analysis of presence-absence data-can contradict each other when applied to a typical microbiome dataset. In our dataset, for example, 14% of species-pairs predicted to be significantly correlated by r were not predicted to be significantly correlated using J, while 37.4% of species-pairs predicted to be significantly correlated by J were not predicted to be significantly correlated using r. Mismatch was particularly common among species-pairs with at least one rare species (<10% prevalence), explaining why r and J might differ more strongly in microbiome datasets, where there are large numbers of rare taxa. Indeed 74% of all species-pairs in our study had at least one rare species. Next, we show how Pearson's correlation coefficient can result in artificial inflation of positive taxon relationships and how this is a particular problem for microbiome studies. We then illustrate how Jaccard's index of similarity (J) can yield improvements over Pearson's correlation coefficient. However, the standard null model for Jaccard's index is flawed, and thus introduces its own set of spurious conclusions. We thus identify a better null model based on a hypergeometric distribution, which appropriately corrects for species prevalence. This model is available from recent statistics literature, and can be used for evaluating the significance of any value of an empirically observed Jaccard's index. The resulting simple, yet effective method for handling correlation analysis of microbial presence-absence datasets provides a robust means of testing and finding relationships and/or shared environmental responses among microbial taxa.

High-dimensional linear state space models for dynamic microbial interaction networks

Medical researchers are increasingly interested in knowing how the complex community of micro-organisms living on human body impacts human health. Key to this is to understand how the microbes interact with each other. Time-course studies on human microbiome indicate that the composition of microbiome changes over short time periods, primarily as a consequence of synergistic and antagonistic interactions of the members of the microbiome with each other and with the environment. Knowledge of the abundance of bacteria-which are the predominant members of the human microbiome-in such time-course studies along with appropriate mathematical models will allow us to identify key dynamic interaction networks within the microbiome. However, the high-dimensional nature of these data poses significant challenges to the development of such mathematical models. We propose a high-dimensional linear State Space Model (SSM) with a new Expectation-Regularization-Maximization (ERM) algorithm to construct a dynamic Microbial Interaction Network (MIN). System noise and measurement noise can be separately specified through SSMs. In order to deal with the problem of high-dimensional parameter space in the SSMs, the proposed new ERM algorithm employs the idea of the adaptive LASSO-based variable selection method so that the sparsity property of MINs can be preserved. We performed simulation studies to evaluate the proposed ERM algorithm for variable selection. The proposed method is applied to identify the dynamic MIN from a time-course vaginal microbiome study of women. This method is amenable to future developments, which may include interactions between microbes and the environment.

A Two-Component System Regulates Bacteroides fragilis Toxin to Maintain Intestinal Homeostasis and Prevent Lethal Disease

Intestinal microbes are recognized for their role in human disease. Enterotoxigenic Bacteroides fragilis (ETBF) has been implicated in inflammatory bowel disease and colorectal cancer; however, colonization alone is insufficient to cause these illnesses. We hypothesized that homeostasis in healthy carriers is maintained by colonic mucus, the major constituent of which is the glycoprotein Muc2. We found that Muc2-deficient mice succumb to lethal disease from ETBF colonization in a B. fragilis toxin (BFT)-dependent manner. We identify a toxin regulator, the two-component system RprXY, which suppresses BFT expression in vitro and in vivo. Overexpression of either component was sufficient to prevent lethal disease in Muc2-deficient mice. Our studies demonstrate that homeostasis in the context of ETBF colonization is dependent on a dynamic interaction between intestinal mucus, a bacterial toxin, and a toxin regulatory system. Regulation of virulence may offer a therapeutic target to maintain intestinal homeostasis in susceptible patients.