Pleiotropy informed adaptive association test of multiple traits using genome‐wide association study summary data

Genetic variants associated with disease outcomes can be used to develop personalized treatment. To reach this precision medicine goal, hundreds of large‐scale genome‐wide association studies (GWAS) have been conducted in the past decade to search for promising genetic variants associated with various traits. They have successfully identified tens of thousands of disease‐related variants. However, in total these identified variants explain only part of the variation for most complex traits. There remain many genetic variants with small effect sizes to be discovered, which calls for the development of (a) GWAS with more samples and more comprehensively genotyped variants, for example, the NHLBI Trans‐Omics for Precision Medicine (TOPMed) Program is planning to conduct whole genome sequencing on over 100 000 individuals; and (b) novel and more powerful statistical analysis methods. The current dominating GWAS analysis approach is the “single trait” association test, despite the fact that many GWAS are conducted in deeply phenotyped cohorts including many correlated and well‐characterized outcomes, which can help improve the power to detect novel variants if properly analyzed, as suggested by increasing evidence that pleiotropy, where a genetic variant affects multiple traits, is the norm in genome‐phenome associations. We aim to develop pleiotropy informed powerful association test methods across multiple traits for GWAS. Since it is generally very hard to access individual‐level GWAS phenotype and genotype data for those existing GWAS, due to privacy concerns and various logistical considerations, we develop rigorous statistical methods for pleiotropy informed adaptive multitrait association test methods that need only summary association statistics publicly available from most GWAS. We first develop a pleiotropy test, which has powerful performance for truly pleiotropic variants but is sensitive to the pleiotropy assumption. We then develop a pleiotropy informed adaptive test that has robust and powerful performance under various genetic models. We develop accurate and efficient numerical algorithms to compute the analytical P‐value for the proposed adaptive test without the need of resampling or permutation. We illustrate the performance of proposed methods through application to joint association test of GWAS meta‐analysis summary data for several glycemic traits. Our proposed adaptive test identified several novel loci missed by individual trait based GWAS meta‐analysis. All the proposed methods are implemented in a publicly available R package.     

Crop genome‐wide association study: a harvest of biological relevance

With the advent of rapid genotyping and next‐generation sequencing technologies, genome‐wide association study (GWAS) has become a routine strategy for decoding genotype–phenotype associations in many species. More than 1000 such studies over the last decade have revealed substantial genotype–phenotype associations in crops and provided unparalleled opportunities to probe functional genomics. Beyond the many ‘hits’ obtained, this review summarizes recent efforts to increase our understanding of the genetic architecture of complex traits by focusing on non‐main effects including epistasis, pleiotropy, and phenotypic plasticity. We also discuss how these achievements and the remaining gaps in our knowledge will guide future studies. Synthetic association is highlighted as leading to false causality, which is prevalent but largely underestimated. Furthermore, validation evidence is appealing for future GWAS, especially in the context of emerging genome‐editing technologies.     

Combining the strengths of inverse-variance weighting and Egger regression in Mendelian randomization using a mixture of regressions model

With the increasing availability of large-scale GWAS summary data on various traits, Mendelian randomization (MR) has become commonly used to infer causality between a pair of traits, an exposure and an outcome. It depends on using genetic variants, typically SNPs, as instrumental variables (IVs). The inverse-variance weighted (IVW) method (with a fixed-effect meta-analysis model) is most powerful when all IVs are valid; however, when horizontal pleiotropy is present, it may lead to biased inference. On the other hand, Egger regression is one of the most widely used methods robust to (uncorrelated) pleiotropy, but it suffers from loss of power. We propose a two-component mixture of regressions to combine and thus take advantage of both IVW and Egger regression; it is often both more efficient (i.e. higher powered) and more robust to pleiotropy (i.e. controlling type I error) than either IVW or Egger regression alone by accounting for both valid and invalid IVs respectively. We propose a model averaging approach and a novel data perturbation scheme to account for uncertainties in model/IV selection, leading to more robust statistical inference for finite samples. Through extensive simulations and applications to the GWAS summary data of 48 risk factor-disease pairs and 63 genetically uncorrelated trait pairs, we showcase that our proposed methods could often control type I error better while achieving much higher power than IVW and Egger regression (and sometimes than several other new/popular MR methods). We expect that our proposed methods will be a useful addition to the toolbox of Mendelian randomization for causal inference.     

Methodological Considerations in Estimation of Phenotype Heritability Using Genome-Wide SNP Data,Illustrated by an Analysis of the Heritability of Height in a Large Sample of African Ancestry Adults

Height has an extremely polygenic pattern of inheritance. Genome-wide association studies (GWAS) have revealed hundreds of common variants that are associated with human height at genome-wide levels of significance. However, only a small fraction of phenotypic variation can be explained by the aggregate of these common variants. In a large study of African-American men and women (n = 14,419), we genotyped and analyzed 966,578 autosomal SNPs across the entire genome using a linear mixed model variance components approach implemented in the program GCTA (Yang et al Nat Genet 2010), and estimated an additive heritability of 44.7% (se: 3.7%) for this phenotype in a sample of evidently unrelated individuals. While this estimated value is similar to that given by Yang et al in their analyses, we remain concerned about two related issues: (1) whether in the complete absence of hidden relatedness, variance components methods have adequate power to estimate heritability when a very large number of SNPs are used in the analysis; and (2) whether estimation of heritability may be biased, in real studies, by low levels of residual hidden relatedness. We addressed the first question in a semi-analytic fashion by directly simulating the distribution of the score statistic for a test of zero heritability with and without low levels of relatedness. The second question was addressed by a very careful comparison of the behavior of estimated heritability for both observed (self-reported) height and simulated phenotypes compared to imputation R² as a function of the number of SNPs used in the analysis. These simulations help to address the important question about whether today''s GWAS SNPs will remain useful for imputing causal variants that are discovered using very large sample sizes in future studies of height, or whether the causal variants themselves will need to be genotyped de novo in order to build a prediction model that ultimately captures a large fraction of the variability of height, and by implication other complex phenotypes. Our overall conclusions are that when study sizes are quite large (5,000 or so) the additive heritability estimate for height is not apparently biased upwards using the linear mixed model; however there is evidence in our simulation that a very large number of causal variants (many thousands) each with very small effect on phenotypic variance will need to be discovered to fill the gap between the heritability explained by known versus unknown causal variants. We conclude that today''s GWAS data will remain useful in the future for causal variant prediction, but that finding the causal variants that need to be predicted may be extremely laborious.     

Heritability in the genome-wide association era

Heritability, the fraction of phenotypic variation explained by genetic variation, has been estimated for many phenotypes in a range of populations, organisms, and time points. The recent development of efficient genotyping and sequencing technology has led researchers to attempt to identify the genetic variants responsible for the genetic component of phenotype directly via GWAS. The gap between the phenotypic variance explained by GWAS results and those estimated from classical heritability methods has been termed the "missing heritability problem". In this work, we examine modern methods for estimating heritability, which use the genotype and sequence data directly. We discuss them in the context of classical heritability methods, the missing heritability problem, and describe their implications for understanding the genetic architecture of complex phenotypes.     

A fast algorithm for Bayesian multi-locus model in genome-wide association studies

Genome-wide association studies (GWAS) have identified a large amount of single-nucleotide polymorphisms (SNPs) associated with complex traits. A recently developed linear mixed model for estimating heritability by simultaneously fitting all SNPs suggests that common variants can explain a substantial fraction of heritability, which hints at the low power of single variant analysis typically used in GWAS. Consequently, many multi-locus shrinkage models have been proposed under a Bayesian framework. However, most use Markov Chain Monte Carlo (MCMC) algorithm, which are time-consuming and challenging to apply to GWAS data. Here, we propose a fast algorithm of Bayesian adaptive lasso using variational inference (BAL-VI). Extensive simulations and real data analysis indicate that our model outperforms the well-known Bayesian lasso and Bayesian adaptive lasso models in accuracy and speed. BAL-VI can complete a simultaneous analysis of a lung cancer GWAS data with ~3400 subjects and ~570,000 SNPs in about half a day.     

Robust Gene-Gene Interaction Analysis in Genome Wide Association Studies

Genome-wide association studies (GWAS) have successfully discovered hundreds of associations between genetic variants and complex traits. Most GWAS have focused on the identification of single variants. It has been shown that most of the variants that were discovered by GWAS could only partially explain disease heritability. The explanation for this missing heritability is generally believed to be gene-gene (GG) or gene-environment (GE) interactions and other structural variants. Generalized multifactor dimensionality reduction (GMDR) has been proven to be reasonably powerful in detecting GG and GE interactions; however, its performance has been found to decline when outlying quantitative traits are present. This paper proposes a robust GMDR estimation method (based on the L-estimator and M-estimator estimation methods) in an attempt to reduce the effects caused by outlying traits. A comparison of robust GMDR with the original MDR based on simulation studies showed the former method to outperform the latter. The performance of robust GMDR is illustrated through a real GWA example consisting of 8,577 samples from the Korean population using the Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) level as a phenotype. Robust GMDR identified the KCNH1 gene to have strong interaction effects with other genes on the function of insulin secretion.     

Mining the LIPG allelic spectrum reveals the contribution of rare and common regulatory variants to HDL cholesterol

Genome-wide association studies (GWAS) have successfully identified loci associated with quantitative traits, such as blood lipids. Deep resequencing studies are being utilized to catalogue the allelic spectrum at GWAS loci. The goal of these studies is to identify causative variants and missing heritability, including heritability due to low frequency and rare alleles with large phenotypic impact. Whereas rare variant efforts have primarily focused on nonsynonymous coding variants, we hypothesized that noncoding variants in these loci are also functionally important. Using the HDL-C gene LIPG as an example, we explored the effect of regulatory variants identified through resequencing of subjects at HDL-C extremes on gene expression, protein levels, and phenotype. Resequencing a portion of the LIPG promoter and 5' UTR in human subjects with extreme HDL-C, we identified several rare variants in individuals from both extremes. Luciferase reporter assays were used to measure the effect of these rare variants on LIPG expression. Variants conferring opposing effects on gene expression were enriched in opposite extremes of the phenotypic distribution. Minor alleles of a common regulatory haplotype and noncoding GWAS SNPs were associated with reduced plasma levels of the LIPG gene product endothelial lipase (EL), consistent with its role in HDL-C catabolism. Additionally, we found that a common nonfunctional coding variant associated with HDL-C (rs2000813) is in linkage disequilibrium with a 5' UTR variant (rs34474737) that decreases LIPG promoter activity. We attribute the gene regulatory role of rs34474737 to the observed association of the coding variant with plasma EL levels and HDL-C. Taken together, the findings show that both rare and common noncoding regulatory variants are important contributors to the allelic spectrum in complex trait loci.     

TATES: Efficient Multivariate Genotype-Phenotype Analysis for Genome-Wide Association Studies

To date, the genome-wide association study (GWAS) is the primary tool to identify genetic variants that cause phenotypic variation. As GWAS analyses are generally univariate in nature, multivariate phenotypic information is usually reduced to a single composite score. This practice often results in loss of statistical power to detect causal variants. Multivariate genotype–phenotype methods do exist but attain maximal power only in special circumstances. Here, we present a new multivariate method that we refer to as TATES (Trait-based Association Test that uses Extended Simes procedure), inspired by the GATES procedure proposed by Li et al (2011). For each component of a multivariate trait, TATES combines p-values obtained in standard univariate GWAS to acquire one trait-based p-value, while correcting for correlations between components. Extensive simulations, probing a wide variety of genotype–phenotype models, show that TATES''s false positive rate is correct, and that TATES''s statistical power to detect causal variants explaining 0.5% of the variance can be 2.5–9 times higher than the power of univariate tests based on composite scores and 1.5–2 times higher than the power of the standard MANOVA. Unlike other multivariate methods, TATES detects both genetic variants that are common to multiple phenotypes and genetic variants that are specific to a single phenotype, i.e. TATES provides a more complete view of the genetic architecture of complex traits. As the actual causal genotype–phenotype model is usually unknown and probably phenotypically and genetically complex, TATES, available as an open source program, constitutes a powerful new multivariate strategy that allows researchers to identify novel causal variants, while the complexity of traits is no longer a limiting factor.     

GPA: A Statistical Approach to Prioritizing GWAS Results by Integrating Pleiotropy and Annotation

Results from Genome-Wide Association Studies (GWAS) have shown that complex diseases are often affected by many genetic variants with small or moderate effects. Identifications of these risk variants remain a very challenging problem. There is a need to develop more powerful statistical methods to leverage available information to improve upon traditional approaches that focus on a single GWAS dataset without incorporating additional data. In this paper, we propose a novel statistical approach, GPA (Genetic analysis incorporating Pleiotropy and Annotation), to increase statistical power to identify risk variants through joint analysis of multiple GWAS data sets and annotation information because: (1) accumulating evidence suggests that different complex diseases share common risk bases, i.e., pleiotropy; and (2) functionally annotated variants have been consistently demonstrated to be enriched among GWAS hits. GPA can integrate multiple GWAS datasets and functional annotations to seek association signals, and it can also perform hypothesis testing to test the presence of pleiotropy and enrichment of functional annotation. Statistical inference of the model parameters and SNP ranking is achieved through an EM algorithm that can handle genome-wide markers efficiently. When we applied GPA to jointly analyze five psychiatric disorders with annotation information, not only did GPA identify many weak signals missed by the traditional single phenotype analysis, but it also revealed relationships in the genetic architecture of these disorders. Using our hypothesis testing framework, statistically significant pleiotropic effects were detected among these psychiatric disorders, and the markers annotated in the central nervous system genes and eQTLs from the Genotype-Tissue Expression (GTEx) database were significantly enriched. We also applied GPA to a bladder cancer GWAS data set with the ENCODE DNase-seq data from 125 cell lines. GPA was able to detect cell lines that are biologically more relevant to bladder cancer. The R implementation of GPA is currently available at http://dongjunchung.github.io/GPA/.     

Integrating comprehensive functional annotations to boost power and accuracy in gene-based association analysis

Gene-based association tests aggregate genotypes across multiple variants for each gene, providing an interpretable gene-level analysis framework for genome-wide association studies (GWAS). Early gene-based test applications often focused on rare coding variants; a more recent wave of gene-based methods, e.g. TWAS, use eQTLs to interrogate regulatory associations. Regulatory variants are expected to be particularly valuable for gene-based analysis, since most GWAS associations to date are non-coding. However, identifying causal genes from regulatory associations remains challenging and contentious. Here, we present a statistical framework and computational tool to integrate heterogeneous annotations with GWAS summary statistics for gene-based analysis, applied with comprehensive coding and tissue-specific regulatory annotations. We compare power and accuracy identifying causal genes across single-annotation, omnibus, and annotation-agnostic gene-based tests in simulation studies and an analysis of 128 traits from the UK Biobank, and find that incorporating heterogeneous annotations in gene-based association analysis increases power and performance identifying causal genes.     

FLAGS: A Flexible and Adaptive Association Test for Gene Sets Using Summary Statistics

Genome-wide association studies (GWAS) have been widely used for identifying common variants associated with complex diseases. Despite remarkable success in uncovering many risk variants and providing novel insights into disease biology, genetic variants identified to date fail to explain the vast majority of the heritability for most complex diseases. One explanation is that there are still a large number of common variants that remain to be discovered, but their effect sizes are generally too small to be detected individually. Accordingly, gene set analysis of GWAS, which examines a group of functionally related genes, has been proposed as a complementary approach to single-marker analysis. Here, we propose a flexible and adaptive test for gene sets (FLAGS), using summary statistics. Extensive simulations showed that this method has an appropriate type I error rate and outperforms existing methods with increased power. As a proof of principle, through real data analyses of Crohn’s disease GWAS data and bipolar disorder GWAS meta-analysis results, we demonstrated the superior performance of FLAGS over several state-of-the-art association tests for gene sets. Our method allows for the more powerful application of gene set analysis to complex diseases, which will have broad use given that GWAS summary results are increasingly publicly available.     

Quantifying Missing Heritability at Known GWAS Loci

Recent work has shown that much of the missing heritability of complex traits can be resolved by estimates of heritability explained by all genotyped SNPs. However, it is currently unknown how much heritability is missing due to poor tagging or additional causal variants at known GWAS loci. Here, we use variance components to quantify the heritability explained by all SNPs at known GWAS loci in nine diseases from WTCCC1 and WTCCC2. After accounting for expectation, we observed all SNPs at known GWAS loci to explain more heritability than GWAS-associated SNPs on average (). For some diseases, this increase was individually significant: for Multiple Sclerosis (MS) () and for Crohn''s Disease (CD) (); all analyses of autoimmune diseases excluded the well-studied MHC region. Additionally, we found that GWAS loci from other related traits also explained significant heritability. The union of all autoimmune disease loci explained more MS heritability than known MS SNPs () and more CD heritability than known CD SNPs (), with an analogous increase for all autoimmune diseases analyzed. We also observed significant increases in an analysis of Rheumatoid Arthritis (RA) samples typed on ImmunoChip, with more heritability from all SNPs at GWAS loci () and more heritability from all autoimmune disease loci () compared to known RA SNPs (including those identified in this cohort). Our methods adjust for LD between SNPs, which can bias standard estimates of heritability from SNPs even if all causal variants are typed. By comparing adjusted estimates, we hypothesize that the genome-wide distribution of causal variants is enriched for low-frequency alleles, but that causal variants at known GWAS loci are skewed towards common alleles. These findings have important ramifications for fine-mapping study design and our understanding of complex disease architecture.