Anne Gatewood Hoen

Introduction: The microbial interactions within the human microbiome are complex, and few methods are available to identify these interactions within a longitudinal microbial abundance framework. Existing methods typically impose restrictive constraints, such as requiring long sequences and equal spacing, on the data format which in many cases are violated.Methods: To identify microbial interaction networks (MINs) with general longitudinal data settings, we propose a stationary Gaussian graphical model (SGGM) based on 16S rRNA gene sequencing data. In the SGGM, data can be arbitrarily spaced, and there are no restrictions on the length of data sequences from a single subject. Based on the SGGM, EM -type algorithms are devised to compute the L1-penalized maximum likelihood estimate of MINs. The algorithms employ the classical graphical LASSO algorithm as the building block and can be implemented efficiently. Results: Extensive simulation studies show that the proposed algorithms can significantly outperform the conventional algorithms if the correlations among the longitudinal data are reasonably high. When the assumptions in the SGGM areviolated, e.g., zero inflation or data from heterogeneous microbial communities, the proposed algorithms still demonstrate robustness and perform better than the other existing algorithms. The algorithms are applied to a 16S rRNA gene sequencing data set from patients with cystic fibrosis. The results demonstrate strong evidence of an association between the MINs and the phylogenetic tree, indicating that the genetically related taxa tend to have more/stronger interactions. These results strengthen the existing findings in literature. Discussion: The proposed algorithms can potentially be used to explore the network structure in genome, metabolome etc. as well.

MicroRNAs (miRNA) in extracellular vesicles and particles (EVPs) in maternal circulation during pregnancy and in human milk postpartum are hypothesized to facilitate maternal–offspring communication via epigenetic regulation. However, factors influencing maternal EVP miRNA profiles during these two critical developmental windows remain largely unknown. In a pilot study of 54 mother–child dyads in the New Hampshire Birth Cohort Study, we profiled 798 EVP miRNAs, using the NanoString nCounter platform, in paired maternal second-trimester plasma and mature (6-week) milk samples. In adjusted models, total EVP miRNA counts were lower for plasma samples collected in the afternoon compared with the morning (p = 0.024). Infant age at sample collection was inversely associated with total miRNA counts in human milk EVPs (p = 0.040). Milk EVP miRNA counts were also lower among participants who were multiparous after delivery (p = 0.047), had a pre-pregnancy BMI > 25 kg/m² (p = 0.037), or delivered their baby via cesarean section (p = 0.021). In post hoc analyses, we also identified 22 specific EVP miRNA that were lower among participants who delivered their baby via cesarean section (Q < 0.05). Target genes of delivery mode-associated miRNAs were over-represented in pathways related to satiety signaling in infants (e.g., CCKR signaling) and mammary gland development and lactation (e.g., FGF signaling, EGF receptor signaling). In conclusion, we identified several key factors that may influence maternal EVP miRNA composition during two critical developmental windows, which should be considered in future studies investigating EVP miRNA roles in maternal and child health.

Antimicrobial resistance is a growing public health burden, but little is known about the effects of antibiotic exposure on the gut resistome. As childhood (0–5 years) represents a sensitive window of microbiome development and a time of relatively high antibiotic use, the aims of this systematic review were to evaluate the effects of antibiotic exposure on the gut resistome of young children and identify knowledge gaps. We searched PubMed, Scopus, Web of Science, and the Cochrane Central Register of Controlled Trials. A PICO framework was developed to determine eligibility criteria. Our main outcomes were the mean or median difference in overall resistance gene load and resistome alpha diversity by antibiotic exposure groups. Bias assessment was completed using RoB 2 and ROBINS-I with quality of evidence assessed via the GRADE criteria. From 4885 records identified, 14 studies (3 randomized controlled trials and 11 observational studies) were included in the qualitative review. Eight studies that included information on antibiotic exposure and overall resistance gene load reported no or positive associations. Inconsistent associations were identified for the nine studies that assessed resistome alpha diversity. We identified three main groups of studies based on study design, location, participants, antibiotic exposures, and indication for antibiotics. Overall, the quality of evidence for our main outcomes was rated low or very low, mainly due to potential bias from the selective of reporting results and confounding. We found evidence that antibiotic exposure is associated with changes to the overall gut resistance gene load of children and may influence the diversity of antimicrobial resistance genes. Given the overall quality of the studies, more research is needed to assess how antibiotics impact the resistome of other populations. Nonetheless, this evidence indicates that the gut resistome is worthwhile to consider for antibiotic prescribing practices.

Background: The human microbiome can contribute to pathogeneses of many complex diseases by mediating disease-leading causal pathways. However, standard mediation analysis methods are not adequate to analyze the microbiome as a mediator due to the excessive number of zero-valued sequencing reads in the data and that the relative abundances have to sum to one. The two main challenges raised by the zero-inflated data structure are: (a) disentangling the mediation effect induced by the point mass at zero; and (b) identifying the observed zero-valued data points that are not zero (i.e., false zeros). Methods: We develop a novel marginal mediation analysis method under the potential-outcomes framework to address the issues. We also show that the marginal model can account for the compositional structure of microbiome data. Results: The mediation effect can be decomposed into two components that are inherent to the two-part nature of zero-inflated distributions. With probabilistic models to account for observing zeros, we also address the challenge with false zeros. A comprehensive simulation study and the application in a real microbiome study showcase our approach in comparison with existing approaches. Conclusions: When analyzing the zero-inflated microbiome composition as the mediators, MarZIC approach has better performance than standard causal mediation analysis approaches and existing competing approach.

Background: Young children are frequently exposed to antibiotics, with the potential for collateral consequences to the gut microbiome. The impact of antibiotic exposures to off-target microbes (i.e., bacteria not targeted by treatment) and antibiotic resistance genes (ARGs) is poorly understood. Methods: We used metagenomic sequencing data from paired stool samples collected prior to antibiotic exposure and at 1 year from over 200 infants and a difference-in-differences approach to assess the relationship between subsequent exposures and the abundance or compositional diversity of microbes and ARGs while adjusting for covariates. Results: By 1 year, the abundance of multiple species and ARGs differed by antibiotic exposure. Compared to infants never exposed to antibiotics, Bacteroides vulgatus relative abundance increased by 1.72% (95% CI: 0.19, 3.24) while Bacteroides fragilis decreased by 1.56% (95% CI: −4.32, 1.21). Bifidobacterium species also exhibited opposing trends. ARGs associated with exposure included class A beta-lactamase gene CfxA6. Among infants attending day care, Escherichia coli and ARG abundance were both positively associated with antibiotic use. Conclusion: Novel findings, including the importance of day care attendance, were identified through considering microbiome data at baseline and post-intervention. Thus, our study design and approach have important implications for future studies evaluating the unintended impacts of antibiotics. Impact: The impact of antibiotic exposure to off-target microbes and antibiotic resistance genes in the gut is poorly defined. We quantified these impacts in two cohort studies using a difference-in-differences approach. Novel to microbiome studies, we used pre/post-antibiotic data to emulate a randomized controlled trial. Compared to infants unexposed to antibiotics between baseline and 1 year, the relative abundance of multiple off-target species and antibiotic resistance genes was altered. Infants who attended day care and were exposed to antibiotics within the first year had a higher abundance of Escherichia coli and antibiotic resistance genes; a novel finding warranting further investigation.

Background: Young children are frequently exposed to antibiotics for otitis media and respiratory infections, with the potential for collateral consequences on the gut microbiome. The impact of antibiotic exposures to off-target microbes (i.e., bacteria not targeted by antibiotic treatment) and antibiotic resistance genes (ARGs) is unknown. Methods: We used metagenomic sequencing data from paired stool samples collected prior to antibiotic exposure and at 1 year from over 200 infants and a difference-in-differences approach to assess the relationship between subsequent exposures and the abundance or compositional diversity of off-target microbes and ARGs while adjusting for covariates. Results: By 1 year, the relative abundance of multiple species and ARGs differed by antibiotic exposure. Compared to infants never exposed to antibiotics, Bacteroides vulgatus relative abundance increased by 1.72% (95%CI:0.19,3.24) while Bacteroides fragilis decreased by 1.56% (95%CI:-4.32,1.21). Bifidobacterium species also exhibited opposing trends suggesting differential antibiotic selection. Overall, antibiotic exposure was associated with a dose-dependent decrease in alpha diversity of off-target microbes. ARGs associated with antibiotic exposure included class A beta-lactamase gene CfxA6 . Among infants attending day care, Escherichia coli and ARG abundance were both positively associated with antibiotic use. Conclusion: Further quantifying impacts to off-target microbes and ARGs has implications for antibiotic stewardship Impact: Infants are frequently exposed to antibiotics for respiratory illnesses, but the extent of impact to off-target gut microbiota and antibiotic resistance genes is unknown. We quantified these impacts in 2 cohort studies using a difference-in-differences approach. Novel to the microbiome space, this enabled us to use pre/post antibiotic data to emulate a randomized controlled trial. Compared to infants unexposed to antibiotics between six weeks and 1 year, the relative abundance of multiple off-target species and antibiotic resistance genes was altered. This research highlights the need to consider the microbiome in antibiotic stewardship and offers a new framework for quantifying change.

Cesarean delivery and formula feeding have both been implicated as important factors associated with perturbations to the infant gut microbiome. To investigate the functional metabolic response of the infant gut microbial milieu to these factors, we profiled the stool metabolomes of 121 infants from a US pregnancy cohort study at approximately 6 weeks of life and evaluated associations with delivery mode and feeding method. Multivariate analysis of six-week stool metabolomic profiles indicated discrimination by both delivery mode and diet. For diet, exclusively breast-fed infants exhibited metabolomic profiles that were distinct from both exclusively formula-fed and combination-fed infants, which were relatively more similar to each other in metabolomic profile. We also identified individual metabolites that were important for differentiating delivery mode groups and feeding groups and metabolic pathways related to delivery mode and feeding type. We conclude based on previous work and this current study that the microbial communities colonizing the gastrointestinal tracts of infants are not only taxonomically, but also functionally distinct when compared according to delivery mode and feeding groups. Further, different sets of metabolites and metabolic pathways define delivery mode and diet metabotypes.

Background: The infant intestinal microbiome plays an important role in metabolism and immune development with impacts on lifelong health. The linkage between the taxonomic composition of the microbiome and its metabolic phenotype is undefined and complicated by redundancies in the taxon-function relationship within microbial communities. To inform a more mechanistic understanding of the relationship between the microbiome and health, we performed an integrative statistical and machine learning-based analysis of microbe taxonomic structure and metabolic function in order to characterize the taxa-function relationship in early life.Results: Stool samples collected from infants enrolled in the New Hampshire Birth Cohort Study (NHBCS) at approximately 6-weeks (n = 158) and 12-months (n = 282) of age were profiled using targeted and untargeted nuclear magnetic resonance (NMR) spectroscopy as well as DNA sequencing of the V4-V5 hypervariable region from the bacterial 16S rRNA gene. There was significant inter-omic concordance based on Procrustes analysis (6 weeks:p = 0.056; 12 months:p = 0.001), however this association was no longer significant when accounting for phylogenetic relationships using generalized UniFrac distance metric (6 weeks:p = 0.376; 12 months:p = 0.069). Sparse canonical correlation analysis showed significant correlation, as well as identifying sets of microbe/metabolites driving microbiome-metabolome relatedness. Performance of machine learning models varied across different metabolites, with support vector machines (radial basis function kernel) being the consistently top ranked model. However, predictive R²values demonstrated poor predictive performance across all models assessed (avg: − 5.06% -- 6 weeks; − 3.7% -- 12 months). Conversely, the Spearman correlation metric was higher (avg: 0.344–6 weeks; 0.265–12 months). This demonstrated that taxonomic relative abundance was not predictive of metabolite concentrations.Conclusions: Our results suggest a degree of overall association between taxonomic profiles and metabolite concentrations. However, lack of predictive capacity for stool metabolic signatures reflects, in part, the possible role of functional redundancy in defining the taxa-function relationship in early life as well as the bidirectional nature of the microbiome-metabolome association. Our results provide evidence in favor of a multi-omic approach for microbiome studies, especially those focused on health outcomes.

The human gastrointestinal tract is an environment that hosts an ecosystem of microorganisms essential to human health. Vital biological processes emerge from fundamental inter- and intra-species molecular interactions that influence the assembly and composition of the gut microbiota ecology. Here we quantify the complexity of the ecological relationships within the human infant gut microbiota ecosystem as a function of the information contained in the nonlinear associations of a sequence of increasingly-specified maximum entropy representations of the system. Our paradigm frames the ecological state, in terms of the presence or absence of individual microbial ecological units that are identified by amplicon sequence variants (ASV) in the gut microenvironment, as a function of both the ecological states of its neighboring units and, in a departure from standard graphical model representations, the associations among the units within its neighborhood. We characterize the order of the system based on the relative quantity of statistical information encoded by high-order statistical associations of the infant gut microbiota.

What is the relationship between mortality and satellite images as elucidated through the use of Convolutional Neural Networks? Background: Following a century of increase, life expectancy in the United States has stagnated and begun to decline in recent decades. Using satellite images and street view images, prior work has demonstrated associations of the built environment with income, education, access to care and health factors such as obesity. However, assessment of learned image feature relationships with variation in crude mortality rate across the United States has been lacking. Objective: We sought to investigate if county-level mortality rates in the U.S. could be predicted from satellite images. Methods: Satellite images of neighborhoods surrounding schools were extracted with the Google Static Maps application programming interface for 430 counties representing approximately 68.9% of the US population. A convolutional neural network was trained using crude mortality rates for each county in 2015 to predict mortality. Learned image features were interpreted using Shapley Additive Feature Explanations, clustered, and compared to mortality and its associated covariate predictors. Results: Predicted mortality from satellite images in a held-out test set of counties was strongly correlated to the true crude mortality rate (Pearson r=0.72). Direct prediction of mortality using a deep learning model across a cross-section of 430 U.S. counties identified key features in the environment (e.g. sidewalks, driveways and hiking trails) associated with lower mortality. Learned image features were clustered, and we identified 10 clusters that were associated with education, income, geographical region, race and age. Conclusions: The application of deep learning techniques to remotely-sensed features of the built environment can serve as a useful predictor of mortality in the United States. Although we identified features that were largely associated with demographic information, future modeling approaches that directly identify image features associated with health-related outcomes have the potential to inform targeted public health interventions.