Bayesian shrinkage estimation of high dimensional causal mediation effects in omics studies

Integrating Multi-Omics with environmental data for precision health: A novel analytic framework and case study on prenatal mercury induced childhood fatty liver disease

BACKGROUND

Precision Health aims to revolutionize disease prevention by leveraging information across multiple omic datasets (multi-omics). However, existing methods generally do not consider personalized environmental risk factors (e.g., environmental pollutants).

OBJECTIVE

To develop and apply a precision health framework which combines multiomic integration (including early, intermediate, and late integration, representing sequential stages at which omics layers are combined for modeling) with mediation approaches (including high-dimensional mediation to identify biomarkers, mediation with latent factors to identify pathways, and integrated/quasi-mediation to identify high-risk subpopulations) to identify novel biomarkers of prenatal mercury induced metabolic dysfunction-associated fatty liver disease (MAFLD), elucidate molecular pathways linking prenatal mercury with MAFLD in children, and identify high-risk children based on integrated exposure and multiomics data.

METHODS

This prospective cohort study used data from 420 mother-child pairs from the Human Early Life Exposome (HELIX) project. Mercury concentrations were determined in maternal or cord blood from pregnancy. Cytokeratin 18 (CK-18; a MAFLD biomarker) and five omics layers (DNA Methylation, gene transcription, microRNA, proteins, and metabolites) were measured in blood in childhood (age 6-10 years).

RESULTS

Each standard deviation increase in prenatal mercury was associated with a 0.11 [95% confidence interval: 0.02-0.21] standard deviation increase in CK-18. High dimensional mediation analysis identified 10 biomarkers linking prenatal mercury and CK-18, including six CpG sites and four transcripts. Mediation with latent factors identified molecular pathways linking mercury and MAFLD, including altered cytokine signaling and hepatic stellate cell activation. Integrated/quasi-mediation identified high risk subgroups of children based on unique combinations of exposure levels, omics profiles (driven by epigenetic markers), and MAFLD.

CONCLUSIONS

Prenatal mercury exposure is associated with elevated liver enzymes in childhood, likely through alterations in DNA methylation and gene expression. Our analytic framework can be applied across many different fields and serve as a resource to help guide future precision health investigations.

A Bayesian nonparametric approach for multiple mediators with applications in mental health studies

Mediation analysis with contemporaneously observed multiple mediators is a significant area of causal inference. Recent approaches for multiple mediators are often based on parametric models and thus may suffer from model misspecification. Also, much of the existing literature either only allow estimation of the joint mediation effect or estimate the joint mediation effect just as the sum of individual mediator effects, ignoring the interaction among the mediators. In this article, we propose a novel Bayesian nonparametric method that overcomes the two aforementioned drawbacks. We model the joint distribution of the observed data (outcome, mediators, treatment, and confounders) flexibly using an enriched Dirichlet process mixture with three levels. We use standardization (g-computation) to compute all possible mediation effects, including pairwise and all other possible interaction among the mediators. We thoroughly explore our method via simulations and apply our method to a mental health data from Wisconsin Longitudinal Study, where we estimate how the effect of births from unintended pregnancies on later life mental depression (CES-D) among the mothers is mediated through lack of self-acceptance and autonomy, employment instability, lack of social participation, and increased family stress. Our method identified significant individual mediators, along with some significant pairwise effects.

High-dimensional causal mediation analysis by partial sum statistic and sample splitting strategy in imaging genetics application

Causal mediation analysis provides a systematic approach to explore the causal role of one or more mediators in the association between exposure and outcome. In omics or imaging data analysis, mediators are often high-dimensional, which brings new statistical challenges. Existing methods either violate causal assumptions or fail in interpretable variable selection. Additionally, mediators are often highly correlated, presenting difficulties in selecting and prioritizing top mediators. To address these issues, we develop a framework using Partial Sum Statistic and Sample Splitting Strategy, namely PS5, for high-dimensional causal mediation analysis. The method provides a powerful global mediation test satisfying causal assumptions, followed by an algorithm to select and prioritize active mediators with quantification of individual mediation contributions. We demonstrate its accurate type I error control, superior statistical power, reduced bias in mediation effect estimation, and accurate mediator selection using extensive simulations of varying levels of effect size, signal sparsity, and mediator correlations. Finally, we apply PS5 to an imaging genetics dataset of chronic obstructive pulmonary disease (COPD) patients ( N =8,897) in the COPDGene study to examine the causal mediation role of lung images ( p =5,810) in the associations between polygenic risk score and lung function and between smoking exposure and lung function, respectively. Both causal mediation analyses successfully estimate the global indirect effect and detect mediating image regions. Collectively, we find a region in the lower lobe of the right lung with a strong and concordant mediation effect for both genetic and environmental exposures. This suggests that targeted treatment toward this region might mitigate the severity of COPD due to genetic and smoking effects.

A new framework for exploratory network mediator analysis in omics data

Omics methods are widely used in basic biology and translational medicine research. More and more omics data are collected to explain the impact of certain risk factors on clinical outcomes. To explain the mechanism of the risk factors, a core question is how to find the genes/proteins/metabolites that mediate their effects on the clinical outcome. Mediation analysis is a modeling framework to study the relationship between risk factors and pathological outcomes, via mediator variables. However, high-dimensional omics data are far more challenging than traditional data: (1) From tens of thousands of genes, can we overcome the curse of dimensionality to reliably select a set of mediators? (2) How do we ensure that the selected mediators are functionally consistent? (3) Many biological mechanisms contain nonlinear effects. How do we include nonlinear effects in the high-dimensional mediation analysis? (4) How do we consider multiple risk factors at the same time? To meet these challenges, we propose a new exploratory mediation analysis framework, medNet, which focuses on finding mediators through predictive modeling. We propose new definitions for predictive exposure, predictive mediator, and predictive network mediator, using a statistical hypothesis testing framework to identify predictive exposures and mediators. Additionally, two heuristic search algorithms are proposed to identify network mediators, essentially subnetworks in the genome-scale biological network that mediate the effects of single or multiple exposures. We applied medNet on a breast cancer data set and a metabolomics data set combined with food intake questionnaire data. It identified functionally consistent network mediators for the exposures' impact on the outcome, facilitating data interpretation.

Methods for mediation analysis with high-dimensional DNA methylation data: Possible choices and comparisons

Epigenetic researchers often evaluate DNA methylation as a potential mediator of the effect of social/environmental exposures on a health outcome. Modern statistical methods for jointly evaluating many mediators have not been widely adopted. We compare seven methods for high-dimensional mediation analysis with continuous outcomes through both diverse simulations and analysis of DNAm data from a large multi-ethnic cohort in the United States, while providing an R package for their seamless implementation and adoption. Among the considered choices, the best-performing methods for detecting active mediators in simulations are the Bayesian sparse linear mixed model (BSLMM) and high-dimensional mediation analysis (HDMA); while the preferred methods for estimating the global mediation effect are high-dimensional linear mediation analysis (HILMA) and principal component mediation analysis (PCMA). We provide guidelines for epigenetic researchers on choosing the best method in practice and offer suggestions for future methodological development.

Doubly robust evaluation of high-dimensional surrogate markers

When evaluating the effectiveness of a treatment, policy, or intervention, the desired measure of efficacy may be expensive to collect, not routinely available, or may take a long time to occur. In these cases, it is sometimes possible to identify a surrogate outcome that can more easily, quickly, or cheaply capture the effect of interest. Theory and methods for evaluating the strength of surrogate markers have been well studied in the context of a single surrogate marker measured in the course of a randomized clinical study. However, methods are lacking for quantifying the utility of surrogate markers when the dimension of the surrogate grows. We propose a robust and efficient method for evaluating a set of surrogate markers that may be high-dimensional. Our method does not require treatment to be randomized and may be used in observational studies. Our approach draws on a connection between quantifying the utility of a surrogate marker and the most fundamental tools of causal inference-namely, methods for robust estimation of the average treatment effect. This connection facilitates the use of modern methods for estimating treatment effects, using machine learning to estimate nuisance functions and relaxing the dependence on model specification. We demonstrate that our proposed approach performs well, demonstrate connections between our approach and certain mediation effects, and illustrate it by evaluating whether gene expression can be used as a surrogate for immune activation in an Ebola study.

A high-dimensional mediation model for a neuroimaging mediator: Integrating clinical, neuroimaging, and neurocognitive data to mitigate late effects in pediatric cancer

Pediatric cancer treatment, especially for brain tumors, can have profound and complicated late effects. With the survival rates increasing because of improved detection and treatment, a more comprehensive understanding of the impact of current treatments on neurocognitive function and brain structure is critically needed. A frontline medulloblastoma clinical trial (SJMB03) has collected data, including treatment, clinical, neuroimaging, and cognitive variables. Advanced methods for modeling and integrating these data are critically needed to understand the mediation pathway from the treatment through brain structure to neurocognitive outcomes. We propose an integrative Bayesian mediation analysis approach to model jointly a treatment exposure, a high-dimensional structural neuroimaging mediator, and a neurocognitive outcome and to uncover the mediation pathway. The high-dimensional imaging-related coefficients are modeled via a binary Ising-Gaussian Markov random field prior (BI-GMRF), addressing the sparsity, spatial dependency, and smoothness and increasing the power to detect brain regions with mediation effects. Numerical simulations demonstrate the estimation accuracy, power, and robustness. For the SJMB03 study, the BI-GMRF method has identified white matter microstructure that is damaged by cancer-directed treatment and impacts late neurocognitive outcomes. The results provide guidance on improving treatment planning to minimize long-term cognitive sequela for pediatric brain tumor patients.

Mediating effect of endocrine hormones on association between per- and polyfluoroalkyl substances exposure and birth size: Findings from sheyang mini birth cohort study

BACKGROUND

Prenatal exposure to per- and polyfluoroalkyl substances (PFAS) has been reported to affect fetus growth, but current results were inconsistent and their mechanism remained unclear.

OBJECTIVES

We aimed to evaluate the associations of prenatal exposure to single and/or multiple PFAS with birth size and to elucidate whether thyroid hormones and reproductive hormones mediate these associations.

METHODS

A total of 1087 mother-newborn pairs from Sheyang Mini Birth Cohort Study were included in the present cross-sectional analysis. 12 PFAS, 5 thyroid hormones and 2 reproductive hormones were measured in cord serum. Multiple linear regression models and Bayesian kernel machine regression (BKMR) models were used to examine the associations of PFAS with either birth size or endocrine hormones. One-at-a-time pairwise mediating effect analysis was applied to estimate the mediating effect of single hormone in the association between individual chemical and birth size. High-dimensional mediation approach including elastic net regularization and Bayesian shrinkage estimation were further performed to reduce exposure dimension and figure out the global mediation effects of joint endocrine hormones.

RESULTS

Perfluorononanoic acid (PFNA) exposure was positively associated to weight for length z score [WLZ, per log10-unit: regression coefficient (β) = 0.26, 95% confidence intervals (CI): 0.04, 0.47] and ponderal index (PI, β = 0.56, 95% CI: 0.09, 1.02), and PFAS mixture results fit by BKMR model showed consistent consequences. High-dimensional mediating analyses revealed that thyroid stimulating hormone (TSH) explained 6.7% of the positive association between PFAS mixtures exposure and PI [Total effect (TE) = 1.499 (0.565, 2.405); Indirect effect (IE) = 0.105 (0.015, 0.231)]. Besides, 7.3% of the PI variance was indirectly explained by 7 endocrine hormones jointly [TE = 0.810 (0.802, 0.819); IE = 0.040 (0.038, 0.041)].

CONCLUSIONS

Prenatal PFAS mixtures exposure, especially PFNA, was positively associated to birth size. Such associations were partly mediated by cord serum TSH.

Combined effects of ambient air pollution and PM2.5 components on renal function and the potential mediation effects of metabolic risk factors in China

Growing evidence links long-term air pollution exposure with renal function. However, little research has been conducted on the combined effects of air pollutant mixture on renal function and multiple mediation effects of metabolic risk factors. This study enrolled 8996 adults without chronic kidney disease (CKD) at baseline from the CHCN-BTH cohort study. Three-year exposure to air pollutants [particulate matter ≤ 2.5 µm (PM_2.5), PM₁₀, PM₁, ozone (O₃), nitrogen dioxide (NO₂), sulfur dioxide (SO₂) and carbon monoxide (CO)] and PM_2.5 components [black carbon (BC), ammonium (NH₄⁺), nitrate (NO₃^-), sulfate (SO₄^2-) and organic matter (OM)] were assessed using well-validated machine learning methods. Linear mixed models were applied to investigate the associations between air pollutants and estimated glomerular filtration rate (eGFR). Quantile G-computation was used to assess the combined effects of pollutant mixtures. Causal mediation analysis and Bayesian mediation analysis were employed to estimate the mediation effects of metabolic risk factors. An interquartile range increases in BC (-0.256, 95 %CI: -0.331, -0.180) and OM (-0.603, 95 %CI: -0.810, -0.397) were significantly associated with eGFR decline; while O₃ (1.151, 95 %CI: 0.813, 1.489), PM₁₀ (0.721, 95 %CI: 0.309, 1.133), NH₄⁺ (0.990, 95 %CI: 0.638, 1.342), and NO₃^- (0.610, 95 %CI: 0.405, 0.815) were associated with higher eGFR. The combined effect of the PM_2.5 component mixture was found to be associated with lower eGFR (-1.147, 95 % CI: -1.456, -0.839), with OM contributing 72.4 % of the negative effect. Univariate mediation analyses showed that high-density lipoprotein (HDL) mediated 7.1 %, 6.9 %, and 6.1 % effects of O₃, BC, and OM, respectively. However, these mediation effects were not significant in Bayesian mediation analysis. These findings suggest the effect of the PM_2.5 component mixture on eGFR decline and the strong contribution of OM. Metabolic risk factors may not mediate the effects of air pollutants. Further study is warranted to clarify the potential mechanisms involved.

A Bayesian joint model for compositional mediation effect selection in microbiome data

Analyzing multivariate count data generated by high-throughput sequencing technology in microbiome research studies is challenging due to the high-dimensional and compositional structure of the data and overdispersion. In practice, researchers are often interested in investigating how the microbiome may mediate the relation between an assigned treatment and an observed phenotypic response. Existing approaches designed for compositional mediation analysis are unable to simultaneously determine the presence of direct effects, relative indirect effects, and overall indirect effects, while quantifying their uncertainty. We propose a formulation of a Bayesian joint model for compositional data that allows for the identification, estimation, and uncertainty quantification of various causal estimands in high-dimensional mediation analysis. We conduct simulation studies and compare our method's mediation effects selection performance with existing methods. Finally, we apply our method to a benchmark data set investigating the sub-therapeutic antibiotic treatment effect on body weight in early-life mice.

meQTL mapping in the GENOA study reveals genetic determinants of DNA methylation in African Americans

Identifying genetic variants that are associated with variation in DNA methylation, an analysis commonly referred to as methylation quantitative trait locus (meQTL) mapping, is an important first step towards understanding the genetic architecture underlying epigenetic variation. Most existing meQTL mapping studies have focused on individuals of European ancestry and are underrepresented in other populations, with a particular absence of large studies in populations with African ancestry. We fill this critical knowledge gap by performing a large-scale cis-meQTL mapping study in 961 African Americans from the Genetic Epidemiology Network of Arteriopathy (GENOA) study. We identify a total of 4,565,687 cis-acting meQTLs in 320,965 meCpGs. We find that 45% of meCpGs harbor multiple independent meQTLs, suggesting potential polygenic genetic architecture underlying methylation variation. A large percentage of the cis-meQTLs also colocalize with cis-expression QTLs (eQTLs) in the same population. Importantly, the identified cis-meQTLs explain a substantial proportion (median = 24.6%) of methylation variation. In addition, the cis-meQTL associated CpG sites mediate a substantial proportion (median = 24.9%) of SNP effects underlying gene expression. Overall, our results represent an important step toward revealing the co-regulation of methylation and gene expression, facilitating the functional interpretation of epigenetic and gene regulation underlying common diseases in African Americans.

A review of statistical strategies to integrate biomarkers of chemical exposure with biomarkers of effect applied in omic-scale environmental epidemiology

Humans are exposed to a growing list of synthetic chemicals, some of them becoming a major public health concern due to their capacity to impact multiple biological endpoints and contribute to a range of chronic diseases. The integration of endogenous (omic) biomarkers of effect in environmental health studies has been growing during the last decade, aiming to gain insight on the potential mechanisms linking the exposures and the clinical conditions. The emergence of high-throughput omic platforms has raised a list of statistical challenges posed by the large dimension and complexity of data generated. Thus, the aim of the present study was to critically review the current state-of-the-science about statistical approaches used to integrate endogenous biomarkers in environmental-health studies linking chemical exposures with health outcomes. The present review specifically focused on internal exposure to environmental chemical pollutants, involving both persistent organic pollutants (POPs), non-persistent pollutants like phthalates or bisphenols, and metals. We identified 42 eligible articles published since 2016, reporting 48 different statistical workflows, mostly focused on POPs and using metabolomic profiling in the intermediate layer. The outcomes were mainly binary and focused on metabolic disorders. A large diversity of statistical strategies were reported to integrate chemical mixtures and endogenous biomarkers to characterize their associations with health conditions. Multivariate regression models were the most predominant statistical method reported in the published workflows, however some studies applied latent based methods or multipollutant models to overcome the specific constraints of omic or exposure of data. A minority of studies used formal mediation analysis to characterize the indirect effects mediated by the endogenous biomarkers. The principles of each specific statistical method and overall workflow set-up are summarized in the light of highlighting their applicability, strengths and weaknesses or interpretability to gain insight into the causal structures underlying the triad: exposure, effect-biomarker and outcome.

Methods for large-scale single mediator hypothesis testing: Possible choices and comparisons

Mediation hypothesis testing for a large number of mediators is challenging due to the composite structure of the null hypothesis,H 0 : α β = 0 ${H}_{0}:\alpha \beta =0$ (α $\alpha $ : effect of the exposure on the mediator after adjusting for confounders;β $\beta $ : effect of the mediator on the outcome after adjusting for exposure and confounders). In this paper, we reviewed three classes of methods for large-scale one at a time mediation hypothesis testing. These methods are commonly used for continuous outcomes and continuous mediators assuming there is no exposure-mediator interaction so that the productα β $\alpha \beta $ has a causal interpretation as the indirect effect. The first class of methods ignores the impact of different structures under the composite null hypothesis, namely, (1)α = 0 , β ≠ 0 $\alpha =0,\beta \ne 0$ ; (2)α ≠ 0 , β = 0 $\alpha \ne 0,\beta =0$ ; and (3)α = β = 0 $\alpha =\beta =0$ . The second class of methods weights the reference distribution under each case of the null to form a mixture reference distribution. The third class constructs a composite test statistic using the three p values obtained under each case of the null so that the reference distribution of the composite statistic is approximatelyU ( 0 , 1 ) $U(0,1)$ . In addition to these existing methods, we developed the Sobel-comp method belonging to the second class, which uses a corrected mixture reference distribution for Sobel's test statistic. We performed extensive simulation studies to compare all six methods belonging to these three classes in terms of the false positive rates (FPRs) under the null hypothesis and the true positive rates under the alternative hypothesis. We found that the second class of methods which uses a mixture reference distribution could best maintain the FPRs at the nominal level under the null hypothesis and had the greatest true positive rates under the alternative hypothesis. We applied all methods to study the mediation mechanism of DNA methylation sites in the pathway from adult socioeconomic status to glycated hemoglobin level using data from the Multi-Ethnic Study of Atherosclerosis (MESA). We provide guidelines for choosing the optimal mediation hypothesis testing method in practice and develop an R package medScan available on the CRAN for implementing all the six methods.

Methods for Mediation Analysis with High-Dimensional DNA Methylation Data: Possible Choices and Comparison

Epigenetic researchers often evaluate DNA methylation as a mediator between social/environmental exposures and disease, but modern statistical methods for jointly evaluating many mediators have not been widely adopted. We compare seven methods for high-dimensional mediation analysis with continuous outcomes through both diverse simulations and analysis of DNAm data from a large national cohort in the United States, while providing an R package for their implementation. Among the considered choices, the best-performing methods for detecting active mediators in simulations are the Bayesian sparse linear mixed model by Song et al. (2020) and high-dimensional mediation analysis by Gao et al. (2019); while the superior methods for estimating the global mediation effect are high-dimensional linear mediation analysis by Zhou et al. (2021) and principal component mediation analysis by Huang and Pan (2016). We provide guidelines for epigenetic researchers on choosing the best method in practice and offer suggestions for future methodological development.

Speeding up interval estimation for R2-based mediation effect of high-dimensional mediators via cross-fitting

Mediation analysis is a useful tool in biomedical research to investigate how molecular phenotypes, such as gene expression, mediate the effect of an exposure on health outcomes. However, commonly used mean-based total mediation effect measures may suffer from cancellation of component-wise mediation effects of opposite directions in the presence of high-dimensional omics mediators. To overcome this limitation, a variance-based R-squared total mediation effect measure has been recently proposed, which, nevertheless, relies on the computationally intensive nonparametric bootstrap for confidence interval estimation. In this work, we formulate a more efficient two-stage cross-fitted estimation procedure for the R-squared measure. To avoid potential bias, we perform iterative Sure Independence Screening (iSIS) in two subsamples to exclude the non-mediators, followed by ordinary least squares (OLS) regressions for the variance estimation. We then construct confidence intervals based on the newly-derived closed-form asymptotic distribution of the R-squared measure. Extensive simulation studies demonstrate that the proposed procedure is hundreds of times more computationally efficient than the resampling-based method with comparable coverage probability. Furthermore, when applied to the Framingham Heart Study, the proposed method replicated the established finding of gene expression mediating age-related variation in systolic blood pressure and discovered the role of gene expression profiles in the relationship between sex and high-density lipoprotein cholesterol. The proposed cross-fitted interval estimation procedure is implemented in R package RsqMed.

Mediation analysis method review of high throughput data

High-throughput technologies have made high-dimensional settings increasingly common, providing opportunities for the development of high-dimensional mediation methods. We aimed to provide useful guidance for researchers using high-dimensional mediation analysis and ideas for biostatisticians to develop it by summarizing and discussing recent advances in high-dimensional mediation analysis. The method still faces many challenges when extended single and multiple mediation analyses to high-dimensional settings. The development of high-dimensional mediation methods attempts to address these issues, such as screening true mediators, estimating mediation effects by variable selection, reducing the mediation dimension to resolve correlations between variables, and utilizing composite null hypothesis testing to test them. Although these problems regarding high-dimensional mediation have been solved to some extent, some challenges remain. First, the correlation between mediators are rarely considered when the variables are selected for mediation. Second, downscaling without incorporating prior biological knowledge makes the results difficult to interpret. In addition, a method of sensitivity analysis for the strict sequential ignorability assumption in high-dimensional mediation analysis is still lacking. An analyst needs to consider the applicability of each method when utilizing them, while a biostatistician could consider extensions and improvements in the methodology.

Multiple mediation effects on association between prenatal triclosan exposure and birth outcomes

BACKGROUND

Triclosan is a broad-spectrum antimicrobial, and was thought to affect intrauterine development, but the mechanism remains unclear.

OBJECTIVE

To explore the association between prenatal triclosan exposure and birth outcomes.

METHODS

Based on 726 mother-child pairs from the Sheyang Mini Birth Cohort Study (SMBCS), we used the available (published) data of triclosan in maternal urines, the hormones including thyroid-related hormones, gonadal hormones in cord blood, and adipokines, trimethylamine-N-oxide (TMAO) and its precursors in cord blood to explore possible health effects of triclosan on birth outcomes through assessing different hormones and parameters, using Bayesian mediation analysis.

RESULTS

Maternal triclosan exposure was associated with ponderal index (β = 0.317) and head circumference (β = -0.172) in generalized linear models. In Bayesian mediation analysis of PI model, estradiol (β = 0.806) and trimethylamine (TMA, β = 0.164) showed positive mediation effects, while total thyroxine (TT4, β = -0.302), leptin (β = -2.023) and TMAO (β = -0.110) showed negative mediation effects. As for model of head circumference, positive mediation effects were observed in free thyroxine (FT4, β = 0.493), TMA (β = 0.178), and TMAO (β = 0.683), negative mediation effects were observed in TT4 (β = -0.231), testosterone (β = -0.331), estradiol (β = -1.153), leptin (β = -2.361), choline (β = -0.169), betaine (β = -0.104), acetyl-L-carnitine (β = -0.773).

CONCLUSION

The results indicated triclosan can affect intrauterine growth by interfering thyroid-related hormones, gonadal hormones, adipokines, TMAO and its precursors.

Persistent organic pollutant exposure contributes to Black/White differences in leukocyte telomere length in the National Health and Nutrition Examination Survey

Despite racial disparities in diseases of aging and premature mortality, non-Hispanic Black Americans tend to have longer leukocyte telomere length (LTL), a biomarker of cellular aging, than non-Hispanic White Americans. Previous findings suggest that exposure to certain persistent organic pollutants (POPs) is both racially-patterned and associated with longer LTL. We examine whether Black/White differences in LTL are explained by differences in exposure to 15 POPs by estimating the indirect effect (IE) of self-reported race on LTL that is mediated through nine polychlorinated biphenyls (PCBs), three furans, and three dioxins, as well as their mixtures. Our study population includes 1,251 adults from the 1999-2000 and 2001-2002 cycles of the cross-sectional National Health and Nutrition Examination Survey. We characterized single-pollutant mediation effects by constructing survey-weighted linear regression models. We also implemented various approaches to quantify a global mediation effect of all POPs, including unpenalized linear regression, ridge regression, and examination of three summary exposure scores. We found support for the hypothesis that exposure to PCBs partially mediates Black/White differences in LTL. In single-pollutant models, there were significant IEs of race on LTL through six individual PCBs (118, 138, 153, 170, 180, and 187). Ridge regression (0.013, CI 0.001, 0.023; 26.0% mediated) and models examining summative exposure scores with linear combinations derived from principal components analysis (0.019, CI 0.009, 0.029; 34.8% mediated) and Toxic Equivalency Quotient (TEQ) scores (0.016, CI 0.005, 0.026; 28.8% mediated) showed significant IEs when incorporating survey weights. Exposures to individual POPs and their mixtures, which may arise from residential and occupational segregation, may help explain why Black Americans have longer LTL than their White counterparts, providing an environmental explanation for counterintuitive race differences in cellular aging.

Bayesian network mediation analysis with application to the brain functional connectome

The brain functional connectome, the collection of interconnected neural circuits along functional networks, facilitates a cutting-edge understanding of brain functioning, and has a potential to play a mediating role within the effect pathway between an exposure and an outcome. While existing mediation analytic approaches are capable of providing insight into complex processes, they mainly focus on a univariate mediator or mediator vector, without considering network-variate mediators. To fill the methodological gap and accomplish this exciting and urgent application, in the article, we propose an integrative mediation analysis under a Bayesian paradigm with networks entailing the mediation effect. To parameterize the network measurements, we introduce individually specified stochastic block models with unknown block allocation, and naturally bridge effect elements through the latent network mediators induced by the connectivity weights across network modules. To enable the identification of truly active mediating components, we simultaneously impose a feature selection across network mediators. We show the superiority of our model in estimating different effect components and selecting active mediating network structures. As a practical illustration of this approach's application to network neuroscience, we characterize the relationship between a therapeutic intervention and opioid abstinence as mediated by brain functional sub-networks.

Improve the Colorectal Cancer Diagnosis Using Gut Microbiome Data

In the United States, colorectal cancer is the second largest cause of cancer death, and accurate early detection and identification of high-risk patients is a high priority. Although fecal screening tests are available, the close relationship between colorectal cancer and the gut microbiome has generated considerable interest. We describe a machine learning method for gut microbiome data to assist in diagnosing colorectal cancer. Our methodology integrates feature engineering, mediation analysis, statistical modeling, and network analysis into a novel unified pipeline. Simulation results illustrate the value of the method in comparison to existing methods. For predicting colorectal cancer in two real datasets, this pipeline showed an 8.7% higher prediction accuracy and 13% higher area under the receiver operator characteristic curve than other published work. Additionally, the approach highlights important colorectal cancer-related taxa for prioritization, such as high levels of Bacteroides fragilis, which can help elucidate disease pathology. Our algorithms and approach can be widely applied for Colorectal cancer prediction using either 16 S rRNA or shotgun metagenomics data.

Multimodal data integration via mediation analysis with high-dimensional exposures and mediators

Motivated by an imaging proteomics study for Alzheimer's disease (AD), in this article, we propose a mediation analysis approach with high-dimensional exposures and high-dimensional mediators to integrate data collected from multiple platforms. The proposed method combines principal component analysis with penalized least squares estimation for a set of linear structural equation models. The former reduces the dimensionality and produces uncorrelated linear combinations of the exposure variables, whereas the latter achieves simultaneous path selection and effect estimation while allowing the mediators to be correlated. Applying the method to the AD data identifies numerous interesting protein peptides, brain regions, and protein-structure-memory paths, which are in accordance with and also supplement existing findings of AD research. Additional simulations further demonstrate the effective empirical performance of the method.

A Bayesian model selection approach to mediation analysis

Genetic studies often seek to establish a causal chain of events originating from genetic variation through to molecular and clinical phenotypes. When multiple phenotypes share a common genetic association, one phenotype may act as an intermediate for the genetic effects on the other. Alternatively, the phenotypes may be causally unrelated but share genetic loci. Mediation analysis represents a class of causal inference approaches used to determine which of these scenarios is most plausible. We have developed a general approach to mediation analysis based on Bayesian model selection and have implemented it in an R package, bmediatR. Bayesian model selection provides a flexible framework that can be tailored to different analyses. Our approach can incorporate prior information about the likelihood of models and the strength of causal effects. It can also accommodate multiple genetic variants or multi-state haplotypes. Our approach reports posterior probabilities that can be useful in interpreting uncertainty among competing models. We compared bmediatR with other popular methods, including the Sobel test, Mendelian randomization, and Bayesian network analysis using simulated data. We found that bmediatR performed as well or better than these alternatives in most scenarios. We applied bmediatR to proteome data from Diversity Outbred (DO) mice, a multi-parent population, and demonstrate the power of mediation with multi-state haplotypes. We also applied bmediatR to data from human cell lines to identify transcripts that are mediated through or are expressed independently from local chromatin accessibility. We demonstrate that Bayesian model selection provides a powerful and versatile approach to identify causal relationships in genetic studies using model organism or human data.

High-Dimensional Mediation Analysis Based on Additive Hazards Model for Survival Data

Mediation analysis has been extensively used to identify potential pathways between exposure and outcome. However, the analytical methods of high-dimensional mediation analysis for survival data are still yet to be promoted, especially for non-Cox model approaches. We propose a procedure including "two-step" variable selection and indirect effect estimation for the additive hazards model with high-dimensional mediators. We first apply sure independence screening and smoothly clipped absolute deviation regularization to select mediators. Then we use the Sobel test and the BH method for indirect effect hypothesis testing. Simulation results demonstrate its good performance with a higher true-positive rate and accuracy, as well as a lower false-positive rate. We apply the proposed procedure to analyze DNA methylation markers mediating smoking and survival time of lung cancer patients in a TCGA (The Cancer Genome Atlas) cohort study. The real data application identifies four mediate CpGs, three of which are newly found.

Bayesian hierarchical models for high-dimensional mediation analysis with coordinated selection of correlated mediators

We consider Bayesian high-dimensional mediation analysis to identify among a large set of correlated potential mediators the active ones that mediate the effect from an exposure variable to an outcome of interest. Correlations among mediators are commonly observed in modern data analysis; examples include the activated voxels within connected regions in brain image data, regulatory signals driven by gene networks in genome data, and correlated exposure data from the same source. When correlations are present among active mediators, mediation analysis that fails to account for such correlation can be suboptimal and may lead to a loss of power in identifying active mediators. Building upon a recent high-dimensional mediation analysis framework, we propose two Bayesian hierarchical models, one with a Gaussian mixture prior that enables correlated mediator selection and the other with a Potts mixture prior that accounts for the correlation among active mediators in mediation analysis. We develop efficient sampling algorithms for both methods. Various simulations demonstrate that our methods enable effective identification of correlated active mediators, which could be missed by using existing methods that assume prior independence among active mediators. The proposed methods are applied to the LIFECODES birth cohort and the Multi-Ethnic Study of Atherosclerosis (MESA) and identified new active mediators with important biological implications.

Bayesian Sparse Mediation Analysis with Targeted Penalization of Natural Indirect Effects

Causal mediation analysis aims to characterize an exposure's effect on an outcome and quantify the indirect effect that acts through a given mediator or a group of mediators of interest. With the increasing availability of measurements on a large number of potential mediators, like the epigenome or the microbiome, new statistical methods are needed to simultaneously accommodate high-dimensional mediators while directly target penalization of the natural indirect effect (NIE) for active mediator identification. Here, we develop two novel prior models for identification of active mediators in high-dimensional mediation analysis through penalizing NIEs in a Bayesian paradigm. Both methods specify a joint prior distribution on the exposure-mediator effect and mediator-outcome effect with either (a) a four-component Gaussian mixture prior or (b) a product threshold Gaussian prior. By jointly modeling the two parameters that contribute to the NIE, the proposed methods enable penalization on their product in a targeted way. Resultant inference can take into account the four-component composite structure underlying the NIE. We show through simulations that the proposed methods improve both selection and estimation accuracy compared to other competing methods. We applied our methods for an in-depth analysis of two ongoing epidemiologic studies: the Multi-Ethnic Study of Atherosclerosis (MESA) and the LIFECODES birth cohort. The identified active mediators in both studies reveal important biological pathways for understanding disease mechanisms.

Penalized mediation models for multivariate data

Statistical methods to integrate multiple layers of data, from exposures to intermediate traits to outcome variables, are needed to guide interpretation of complex data sets for which variables are likely contributing in a causal pathway from exposure to outcome. Statistical mediation analysis based on structural equation models provide a general modeling framework, yet they can be difficult to apply to high-dimensional data and they are not automated to select the best fitting model. To overcome these limitations, we developed novel algorithms and software to simultaneously evaluate multiple exposure variables, multiple intermediate traits, and multiple outcome variables. Our penalized mediation models are computationally efficient and simulations demonstrate that they produce reliable results for large data sets. Application of our methods to a study of vascular disease demonstrates their utility to identify novel direct effects of single-nucleotide polymorphisms (SNPs) on coronary heart disease and peripheral artery disease, while disentangling the effects of SNPs on the intermediate risk factors including lipids, cigarette smoking, systolic blood pressure, and type 2 diabetes.

Estimation of total mediation effect for high-dimensional omics mediators

BACKGROUND

Environmental exposures can regulate intermediate molecular phenotypes, such as gene expression, by different mechanisms and thereby lead to various health outcomes. It is of significant scientific interest to unravel the role of potentially high-dimensional intermediate phenotypes in the relationship between environmental exposure and traits. Mediation analysis is an important tool for investigating such relationships. However, it has mainly focused on low-dimensional settings, and there is a lack of a good measure of the total mediation effect. Here, we extend an R-squared (R[Formula: see text]) effect size measure, originally proposed in the single-mediator setting, to the moderate- and high-dimensional mediator settings in the mixed model framework.

RESULTS

Based on extensive simulations, we compare our measure and estimation procedure with several frequently used mediation measures, including product, proportion, and ratio measures. Our R[Formula: see text]-based second-moment measure has small bias and variance under the correctly specified model. To mitigate potential bias induced by non-mediators, we examine two variable selection procedures, i.e., iterative sure independence screening and false discovery rate control, to exclude the non-mediators. We establish the consistency of the proposed estimation procedures and introduce a resampling-based confidence interval. By applying the proposed estimation procedure, we found that 38% of the age-related variations in systolic blood pressure can be explained by gene expression profiles in the Framingham Heart Study of 1711 individuals. An R package "RsqMed" is available on CRAN.

CONCLUSION

R-squared (R[Formula: see text]) is an effective and efficient measure for total mediation effect especially under high-dimensional setting.

IUSMMT: Survival mediation analysis of gene expression with multiple DNA methylation exposures and its application to cancers of TCGA

Effective and powerful survival mediation models are currently lacking. To partly fill such knowledge gap, we particularly focus on the mediation analysis that includes multiple DNA methylations acting as exposures, one gene expression as the mediator and one survival time as the outcome. We proposed IUSMMT (intersection-union survival mixture-adjusted mediation test) to effectively examine the existence of mediation effect by fitting an empirical three-component mixture null distribution. With extensive simulation studies, we demonstrated the advantage of IUSMMT over existing methods. We applied IUSMMT to ten TCGA cancers and identified multiple genes that exhibited mediating effects. We further revealed that most of the identified regions, in which genes behaved as active mediators, were cancer type-specific and exhibited a full mediation from DNA methylation CpG sites to the survival risk of various types of cancers. Overall, IUSMMT represents an effective and powerful alternative for survival mediation analysis; our results also provide new insights into the functional role of DNA methylation and gene expression in cancer progression/prognosis and demonstrate potential therapeutic targets for future clinical practice.

Roles of allostatic load, lifestyle and clinical risk factors in mediating the association between education and coronary heart disease risk in Europe

BACKGROUND

Previous studies have shown that differential exposure to lifestyle factors may mediate the association between education and coronary heart diseases (CHD). However, few studies have examined the potential roles of allostatic load (AL) or differential susceptibility.

METHODS

25 310 men and 26 018 women aged 35-74 and CHD free at baseline were identified from 21 European cohorts and followed for a median of 10 years, to investigate the mediating role of AL, as well as of smoking, alcohol use and body mass index (BMI), on educational differences in CHD incidence, applying marginal structural models and three-way decomposition.

RESULTS

AL is a mediator of the association between educational status and CHD incidence, with the highest proportion mediated observed among women and largely attributable to differential exposure, (28% (95% CI 19% to 44%)), with 8% (95% CI 0% to 16%) attributable to differential susceptibility. The mediating effects of smoking, alcohol and BMI, compared with AL, were relatively small for both men and women.

CONCLUSION

Overall, the educational inequalities in CHD incidence were partially mediated through differential exposure to AL. By contrast, the mediation of the educational gradient in CHD by investigated lifestyle risk factors was limited. As differential susceptibility in men was found to have a predominant role in the accumulation of AL in low educational classes, the investigation of AL-related risk factors is warranted.

Statistical methods for mediation analysis in the era of high-throughput genomics: Current successes and future challenges

Mediation analysis investigates the intermediate mechanism through which an exposure exerts its influence on the outcome of interest. Mediation analysis is becoming increasingly popular in high-throughput genomics studies where a common goal is to identify molecular-level traits, such as gene expression or methylation, which actively mediate the genetic or environmental effects on the outcome. Mediation analysis in genomics studies is particularly challenging, however, thanks to the large number of potential mediators measured in these studies as well as the composite null nature of the mediation effect hypothesis. Indeed, while the standard univariate and multivariate mediation methods have been well-established for analyzing one or multiple mediators, they are not well-suited for genomics studies with a large number of mediators and often yield conservative p-values and limited power. Consequently, over the past few years many new high-dimensional mediation methods have been developed for analyzing the large number of potential mediators collected in high-throughput genomics studies. In this work, we present a thorough review of these important recent methodological advances in high-dimensional mediation analysis. Specifically, we describe in detail more than ten high-dimensional mediation methods, focusing on their motivations, basic modeling ideas, specific modeling assumptions, practical successes, methodological limitations, as well as future directions. We hope our review will serve as a useful guidance for statisticians and computational biologists who develop methods of high-dimensional mediation analysis as well as for analysts who apply mediation methods to high-throughput genomics studies.

Application of an analytical framework for multivariate mediation analysis of environmental data

Diverse toxicological mechanisms may mediate the impact of environmental toxicants (phthalates, phenols, polycyclic aromatic hydrocarbons, and metals) on pregnancy outcomes. In this study, we introduce an analytical framework for multivariate mediation analysis to identify mediation pathways (q = 61 mediators) in the relationship between environmental toxicants (p = 38 analytes) and gestational age at delivery. Our analytical framework includes: (1) conducting pairwise mediation for unique exposure-mediator combinations, (2) exposure dimension reduction by estimating environmental risk scores, and (3) multivariate mediator analysis using either Bayesian shrinkage mediation analysis, population value decomposition, or mediation pathway penalization. Dimension reduction demonstrates that a one-unit increase in phthalate risk score is associated with a total effect of 1.07 lower gestational age (in weeks) at delivery (95% confidence interval: 0.48-1.67) and eicosanoids from the cytochrome p450 pathway mediated 26% of this effect (95% confidence interval: 4-63%). Eicosanoid products derived from the cytochrome p450 pathway may be important mediators of phthalate toxicity.