Semiparametric Allelic Tests for Mapping Multiple Phenotypes: Binomial Regression and Mahalanobis Distance

A new approach to testing mediation of the microbiome at both the community and individual taxon levels

MOTIVATION

Understanding whether and which microbes played a mediating role between an exposure and a disease outcome are essential for researchers to develop clinical interventions to treat the disease by modulating the microbes. Existing methods for mediation analysis of the microbiome are often limited to a global test of community-level mediation or selection of mediating microbes without control of the false discovery rate (FDR). Further, while the null hypothesis of no mediation at each microbe is a composite null that consists of three types of null, most existing methods treat the microbes as if they were all under the same type of null, leading to excessive false positive results.

RESULTS

We propose a new approach based on inverse regression that regresses the microbiome data at each taxon on the exposure and the exposure-adjusted outcome. Then, the P-values for testing the coefficients are used to test mediation at both the community and individual taxon levels. This approach fits nicely into our Linear Decomposition Model (LDM) framework, so our new method LDM-med, implemented in the LDM framework, enjoys all the features of the LDM, e.g. allowing an arbitrary number of taxa to be tested simultaneously, supporting continuous, discrete, or multivariate exposures and outcomes (including survival outcomes), and so on. Using extensive simulations, we showed that LDM-med always preserved the FDR of testing individual taxa and had adequate sensitivity; LDM-med always controlled the type I error of the global test and had compelling power over existing methods. The flexibility of LDM-med for a variety of mediation analyses is illustrated by an application to a murine microbiome dataset, which identified several plausible mediating taxa.

AVAILABILITY AND IMPLEMENTATION

Our new method has been added to our R package LDM, which is available on GitHub at https://github.com/yijuanhu/LDM.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Extension of PERMANOVA to Testing the Mediation Effect of the Microbiome

Recently, we have seen a growing volume of evidence linking the microbiome and human diseases or clinical outcomes, as well as evidence linking the microbiome and environmental exposures. Now comes the time to assess whether the microbiome mediates the effects of exposures on the outcomes, which will enable researchers to develop interventions to modulate outcomes by modifying microbiome compositions. Use of distance matrices is a popular approach to analyzing complex microbiome data that are high-dimensional, sparse, and compositional. However, the existing distance-based methods for mediation analysis of microbiome data, MedTest and MODIMA, only work well in limited scenarios. PERMANOVA is currently the most commonly used distance-based method for testing microbiome associations. Using the idea of inverse regression, here we extend PERMANOVA to test microbiome-mediation effects by including both the exposure and the outcome as covariates and basing the test on the product of their F statistics. This extension of PERMANOVA, which we call PERMANOVA-med, naturally inherits all the flexible features of PERMANOVA, e.g., allowing adjustment of confounders, accommodating continuous, binary, and multivariate exposure and outcome variables including survival outcomes, and providing an omnibus test that combines the results from analyzing multiple distance matrices. Our extensive simulations indicated that PERMANOVA-med always controlled the type I error and had compelling power over MedTest and MODIMA. Frequently, MedTest had diminished power and MODIMA had inflated type I error. Using real data on melanoma immunotherapy response, we demonstrated the wide applicability of PERMANOVA-med through 16 different mediation analyses, only 6 of which could be performed by MedTest and 4 by MODIMA.

Testing cell-type-specific mediation effects in genome-wide epigenetic studies

Epigenome-wide mediation analysis aims to identify DNA methylation CpG sites that mediate the causal effects of genetic/environmental exposures on health outcomes. However, DNA methylations in the peripheral blood tissues are usually measured at the bulk level based on a heterogeneous population of white blood cells. Using the bulk level DNA methylation data in mediation analysis might cause confounding bias and reduce study power. Therefore, it is crucial to get fine-grained results by detecting mediation CpG sites in a cell-type-specific way. However, there is a lack of methods and software to achieve this goal. We propose a novel method (Mediation In a Cell-type-Specific fashion, MICS) to identify cell-type-specific mediation effects in genome-wide epigenetic studies using only the bulk-level DNA methylation data. MICS follows the standard mediation analysis paradigm and consists of three key steps. In step1, we assess the exposure-mediator association for each cell type; in step 2, we assess the mediator-outcome association for each cell type; in step 3, we combine the cell-type-specific exposure-mediator and mediator-outcome associations using a multiple testing procedure named MultiMed [Sampson JN, Boca SM, Moore SC, et al. FWER and FDR control when testing multiple mediators. Bioinformatics 2018;34:2418-24] to identify significant CpGs with cell-type-specific mediation effects. We conduct simulation studies to demonstrate that our method has correct FDR control. We also apply the MICS procedure to the Normative Aging Study and identify nine DNA methylation CpG sites in the lymphocytes that might mediate the effect of cigarette smoking on the lung function.

A two-step approach to testing overall effect of gene-environment interaction for multiple phenotypes

MOTIVATION

While gene-environment (GxE) interactions contribute importantly to many different phenotypes, detecting such interactions requires well-powered studies and has proven difficult. To address this, we combine two approaches to improve GxE power: simultaneously evaluating multiple phenotypes and using a two-step analysis approach. Previous work shows that the power to identify a main genetic effect can be improved by simultaneously analyzing multiple related phenotypes. For a univariate phenotype, two-step methods produce higher power for detecting a GxE interaction compared to single step analysis. Therefore, we propose a two-step approach to test for an overall GxE effect for multiple phenotypes.

RESULTS

Using simulations we demonstrate that, when more than one phenotype has GxE effect (i.e., GxE pleiotropy), our approach offers substantial gain in power (18%-43%) to detect an aggregate-level GxE effect for a multivariate phenotype compared to an analogous two-step method to identify GxE effect for a univariate phenotype. We applied the proposed approach to simultaneously analyze three lipids, LDL, HDL and Triglyceride with the frequency of alcohol consumption as environmental factor in the UK Biobank. The method identified two loci with an overall GxE effect on the vector of lipids, one of which was missed by the competing approaches.

AVAILABILITY

We provide an R package MPGE implementing the proposed approach which is available from CRAN: https://cran.r-project.org/web/packages/MPGE/index.html.

Effect of non-normality and low count variants on cross-phenotype association tests in GWAS

Many complex human diseases, such as type 2 diabetes, are characterized by multiple underlying traits/phenotypes that have substantially shared genetic architecture. Multivariate analysis of correlated traits has the potential to increase the power of detecting underlying common genetic loci. Several cross-phenotype association methods have been proposed-some require individual-level data on traits and genotypes, while the others require only summary-level data. In this article, we explore whether non-normality of multivariate trait distribution affects the inference from some of the existing multi-trait methods and how that effect is dependent on the allele count of the genetic variant being tested. We find that most of these tests are susceptible to biases that lead to spurious association signals. Even after controlling for confounders that may contribute to non-normality and then applying inverse normal transformation on the residuals of each trait, these tests may have inflated type I errors for variants with low minor allele counts (MACs). A likelihood ratio test of association based on the ordinal regression of individual-level genotype conditional on the traits seems to be the least biased and can maintain type I error when the MAC is reasonably large (e.g., MAC > 30). Application of these methods to publicly available summary statistics of eight amino acid traits on European samples seem to exhibit systematic inflation (especially for variants with low MAC), which is consistent with our findings from simulation experiments.

Joint analysis of multiple phenotypes in association studies using allele-based clustering approach for non-normal distributions

In the study of complex diseases, several correlated phenotypes are usually measured. There is also increasing evidence showing that testing the association between a single-nucleotide polymorphism (SNP) and multiple-dependent phenotypes jointly is often more powerful than analyzing only one phenotype at a time. Therefore, developing statistical methods to test for genetic association with multiple phenotypes has become increasingly important. In this paper, we develop an Allele-based Clustering Approach (ACA) for the joint analysis of multiple non-normal phenotypes in association studies. In ACA, we consider the alleles at a SNP of interest as a dependent variable with two classes, and the correlated phenotypes as predictors to predict the alleles at the SNP of interest. We perform extensive simulation studies to evaluate the performance of ACA and compare the power of ACA with the powers of Adaptive Fisher's Combination test, Trait-based Association Test that uses Extended Simes procedure, Fisher's Combination test, the standard MANOVA, and the joint model of Multiple Phenotypes. Our simulation studies show that the proposed method has correct type I error rates and is much more powerful than other methods for some non-normal distributions.

Enabling genome-wide association testing with multiple diseases and no healthy controls

MOTIVATION

While large-scale whole genome sequencing is feasible the high costs compel investigators to focus on disease subjects. As a result large sequencing datasets of samples with different diseases are often readily available, but not healthy controls to contrast them with. While it is possible to perform an association study using only diseases, the associations could be driven by a disease acting as a control and not the focal disease.

METHODS

We developed a genotype-on-phenotype reverse regression with a Bayesian spike and slab prior to enable association testing in datasets with multiple diseases. This method, referred to as revreg, flagged associations (both common and rare) that were driven by diseases that were not of primary interest.

RESULTS

Based on simulations, revreg had 80% power to detect an odds ratio of 1.74 for common variants (3500 samples total) and 3.73 for rare variants (14,000 samples total), with minimal type I error. For common variants, we tested this method on 3657 whole genome sequenced samples aimed at discovering variants associated with disease risk of Chronic Obstructive Pulmonary Disease using three other diseases as controls. We demonstrated detection of six highly significant associations likely due to Age-Related Macular Degeneration. In an exome dataset of 8836 samples aimed at characterizing rare variants associated with disease risk of Asthma, using five other diseases as controls, we detected and removed genic regions due to AMD (C3, CFH, CFHR5, CFI, and DNMT3A) and RA (KRTAP13-4).

Statistical methods to detect pleiotropy in human complex traits

In recent years pleiotropy, the phenomenon of one genetic locus influencing several traits, has become a widely researched field in human genetics. With the increasing availability of genome-wide association study summary statistics, as well as the establishment of deeply phenotyped sample collections, it is now possible to systematically assess the genetic overlap between multiple traits and diseases. In addition to increasing power to detect associated variants, multi-trait methods can also aid our understanding of how different disorders are aetiologically linked by highlighting relevant biological pathways. A plethora of available tools to perform such analyses exists, each with their own advantages and limitations. In this review, we outline some of the currently available methods to conduct multi-trait analyses. First, we briefly introduce the concept of pleiotropy and outline the current landscape of pleiotropy research in human genetics; second, we describe analytical considerations and analysis methods; finally, we discuss future directions for the field.

A novel method to test associations between a weighted combination of phenotypes and genetic variants

Many complex diseases like diabetes, hypertension, metabolic syndrome, et cetera, are measured by multiple correlated phenotypes. However, most genome-wide association studies (GWAS) focus on one phenotype of interest or study multiple phenotypes separately for identifying genetic variants associated with complex diseases. Analyzing one phenotype or the related phenotypes separately may lose power due to ignoring the information obtained by combining phenotypes, such as the correlation between phenotypes. In order to increase statistical power to detect genetic variants associated with complex diseases, we develop a novel method to test a weighted combination of multiple phenotypes (WCmulP). We perform extensive simulation studies as well as real data (COPDGene) analysis to evaluate the performance of the proposed method. Our simulation results show that WCmulP has correct type I error rates and is either the most powerful test or comparable to the most powerful test among the methods we compared. WCmulP also has an outstanding performance for identifying single-nucleotide polymorphisms (SNPs) associated with COPD-related phenotypes.

Methods for meta-analysis of multiple traits using GWAS summary statistics

Genome-wide association studies (GWAS) for complex diseases have focused primarily on single-trait analyses for disease status and disease-related quantitative traits. For example, GWAS on risk factors for coronary artery disease analyze genetic associations of plasma lipids such as total cholesterol, LDL-cholesterol, HDL-cholesterol, and triglycerides (TGs) separately. However, traits are often correlated and a joint analysis may yield increased statistical power for association over multiple univariate analyses. Recently several multivariate methods have been proposed that require individual-level data. Here, we develop metaUSAT (where USAT is unified score-based association test), a novel unified association test of a single genetic variant with multiple traits that uses only summary statistics from existing GWAS. Although the existing methods either perform well when most correlated traits are affected by the genetic variant in the same direction or are powerful when only a few of the correlated traits are associated, metaUSAT is designed to be robust to the association structure of correlated traits. metaUSAT does not require individual-level data and can test genetic associations of categorical and/or continuous traits. One can also use metaUSAT to analyze a single trait over multiple studies, appropriately accounting for overlapping samples, if any. metaUSAT provides an approximate asymptotic P-value for association and is computationally efficient for implementation at a genome-wide level. Simulation experiments show that metaUSAT maintains proper type-I error at low error levels. It has similar and sometimes greater power to detect association across a wide array of scenarios compared to existing methods, which are usually powerful for some specific association scenarios only. When applied to plasma lipids summary data from the METSIM and the T2D-GENES studies, metaUSAT detected genome-wide significant loci beyond the ones identified by univariate analyses. Evidence from larger studies suggest that the variants additionally detected by our test are, indeed, associated with lipid levels in humans. In summary, metaUSAT can provide novel insights into the genetic architecture of a common disease or traits.

A quadratically regularized functional canonical correlation analysis for identifying the global structure of pleiotropy with NGS data

Investigating the pleiotropic effects of genetic variants can increase statistical power, provide important information to achieve deep understanding of the complex genetic structures of disease, and offer powerful tools for designing effective treatments with fewer side effects. However, the current multiple phenotype association analysis paradigm lacks breadth (number of phenotypes and genetic variants jointly analyzed at the same time) and depth (hierarchical structure of phenotype and genotypes). A key issue for high dimensional pleiotropic analysis is to effectively extract informative internal representation and features from high dimensional genotype and phenotype data. To explore correlation information of genetic variants, effectively reduce data dimensions, and overcome critical barriers in advancing the development of novel statistical methods and computational algorithms for genetic pleiotropic analysis, we proposed a new statistic method referred to as a quadratically regularized functional CCA (QRFCCA) for association analysis which combines three approaches: (1) quadratically regularized matrix factorization, (2) functional data analysis and (3) canonical correlation analysis (CCA). Large-scale simulations show that the QRFCCA has a much higher power than that of the ten competing statistics while retaining the appropriate type 1 errors. To further evaluate performance, the QRFCCA and ten other statistics are applied to the whole genome sequencing dataset from the TwinsUK study. We identify a total of 79 genes with rare variants and 67 genes with common variants significantly associated with the 46 traits using QRFCCA. The results show that the QRFCCA substantially outperforms the ten other statistics.

Determining Which Phenotypes Underlie a Pleiotropic Signal

Discovering pleiotropic loci is important to understand the biological basis of seemingly distinct phenotypes. Most methods for assessing pleiotropy only test for the overall association between genetic variants and multiple phenotypes. To determine which specific traits are pleiotropic, we evaluate via simulation and application three different strategies. The first is model selection techniques based on the inverse regression of genotype on phenotypes. The second is a subset-based meta analysis ASSET [Bhattacharjee et al., ], which provides an optimal subset of nonnull traits. And the third is a modified Benjamini-Hochberg (B-H) procedure of controlling the expected false discovery rate [Benjamini and Hochberg, ] in the framework of phenome-wide association study. From our simulations we see that an inverse regression-based approach MultiPhen [O'Reilly et al., ] is more powerful than ASSET for detecting overall pleiotropic association, except for when all the phenotypes are associated and have genetic effects in the same direction. For determining which specific traits are pleiotropic, the modified B-H procedure performs consistently better than the other two methods. The inverse regression-based selection methods perform competitively with the modified B-H procedure only when the phenotypes are weakly correlated. The efficiency of ASSET is observed to lie below and in between the efficiency of the other two methods when the traits are weakly and strongly correlated, respectively. In our application to a large GWAS, we find that the modified B-H procedure also performs well, indicating that this may be an optimal approach for determining the traits underlying a pleiotropic signal.