Methods and results from the genome-wide association group at GAW20

Background

This paper summarizes the contributions from the Genome-wide Association Study group (GWAS group) of the GAW20. The GWAS group contributions focused on topics such as association tests, phenotype imputation, and application of empirical kinships. The goals of the GWAS group contributions were varied. A real or a simulated data set based on the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) study was employed by different methods. Different outcomes and covariates were considered, and quality control procedures varied throughout the contributions.

Results

The consideration of heritability and family structure played a major role in some contributions. The inclusion of family information and adaptive weights based on data were found to improve power in genome-wide association studies. It was proven that gene-level approaches are more powerful than single-marker analysis. Other contributions focused on the comparison between pedigree-based kinship and empirical kinship matrices, and investigated similar results in heritability estimation, association mapping, and genomic prediction. A new approach for linkage mapping of triglyceride levels was able to identify a novel linkage signal.

Conclusions

This summary paper reports on promising statistical approaches and findings of the members of the GWAS group applied on real and simulated data which encompass the current topics of epigenetic and pharmacogenomics.

     

An evolutionary framework for association testing in resequencing studies

Sequencing technologies are becoming cheap enough to apply to large numbers of study participants and promise to provide new insights into human phenotypes by bringing to light rare and previously unknown genetic variants. We develop a new framework for the analysis of sequence data that incorporates all of the major features of previously proposed approaches, including those focused on allele counts and allele burden, but is both more general and more powerful. We harness population genetic theory to provide prior information on effect sizes and to create a pooling strategy for information from rare variants. Our method, EMMPAT (Evolutionary Mixed Model for Pooled Association Testing), generates a single test per gene (substantially reducing multiple testing concerns), facilitates graphical summaries, and improves the interpretation of results by allowing calculation of attributable variance. Simulations show that, relative to previously used approaches, our method increases the power to detect genes that affect phenotype when natural selection has kept alleles with large effect sizes rare. We demonstrate our approach on a population-based re-sequencing study of association between serum triglycerides and variation in ANGPTL4.     

A Protein Domain and Family Based Approach to Rare Variant Association Analysis

BackgroundIt has become common practice to analyse large scale sequencing data with statistical approaches based around the aggregation of rare variants within the same gene. We applied a novel approach to rare variant analysis by collapsing variants together using protein domain and family coordinates, regarded to be a more discrete definition of a biologically functional unit.MethodsUsing Pfam definitions, we collapsed rare variants (Minor Allele Frequency ≤ 1%) together in three different ways 1) variants within single genomic regions which map to individual protein domains 2) variants within two individual protein domain regions which are predicted to be responsible for a protein-protein interaction 3) all variants within combined regions from multiple genes responsible for coding the same protein domain (i.e. protein families). A conventional collapsing analysis using gene coordinates was also undertaken for comparison. We used UK10K sequence data and investigated associations between regions of variants and lipid traits using the sequence kernel association test (SKAT).ResultsWe observed no strong evidence of association between regions of variants based on Pfam domain definitions and lipid traits. Quantile-Quantile plots illustrated that the overall distributions of p-values from the protein domain analyses were comparable to that of a conventional gene-based approach. Deviations from this distribution suggested that collapsing by either protein domain or gene definitions may be favourable depending on the trait analysed.ConclusionWe have collapsed rare variants together using protein domain and family coordinates to present an alternative approach over collapsing across conventionally used gene-based regions. Although no strong evidence of association was detected in these analyses, future studies may still find value in adopting these approaches to detect previously unidentified association signals.     

Comprehensive approach to analyzing rare genetic variants

Recent findings suggest that rare variants play an important role in both monogenic and common diseases. Due to their rarity, however, it remains unclear how to appropriately analyze the association between such variants and disease. A common approach entails combining rare variants together based on a priori information and analyzing them as a single group. Here one must make some assumptions about what to aggregate. Instead, we propose two approaches to empirically determine the most efficient grouping of rare variants. The first considers multiple possible groupings using existing information. The second is an agnostic "step-up" approach that determines an optimal grouping of rare variants analytically and does not rely on prior information. To evaluate these approaches, we undertook a simulation study using sequence data from genes in the one-carbon folate metabolic pathway. Our results show that using prior information to group rare variants is advantageous only when information is quite accurate, but the step-up approach works well across a broad range of plausible scenarios. This agnostic approach allows one to efficiently analyze the association between rare variants and disease while avoiding assumptions required by other approaches for grouping such variants.     

Utilizing Population Controls in Rare-Variant Case-Parent Association Tests

There is great interest in detecting associations between human traits and rare genetic variation. To address the low power implicit in single-locus tests of rare genetic variants, many rare-variant association approaches attempt to accumulate information across a gene, often by taking linear combinations of single-locus contributions to a statistic. Using the right linear combination is key—an optimal test will up-weight true causal variants, down-weight neutral variants, and correctly assign the direction of effect for causal variants. Here, we propose a procedure that exploits data from population controls to estimate the linear combination to be used in an case-parent trio rare-variant association test. Specifically, we estimate the linear combination by comparing population control allele frequencies with allele frequencies in the parents of affected offspring. These estimates are then used to construct a rare-variant transmission disequilibrium test (rvTDT) in the case-parent data. Because the rvTDT is conditional on the parents’ data, using parental data in estimating the linear combination does not affect the validity or asymptotic distribution of the rvTDT. By using simulation, we show that our new population-control-based rvTDT can dramatically improve power over rvTDTs that do not use population control information across a wide variety of genetic architectures. It also remains valid under population stratification. We apply the approach to a cohort of epileptic encephalopathy (EE) trios and find that dominant (or additive) inherited rare variants are unlikely to play a substantial role within EE genes previously identified through de novo mutation studies.     

Weighted Score Tests Implementing Model-Averaging Schemes in Detection of Rare Variants in Case-Control Studies

Multi-locus effect modeling is a powerful approach for detection of genes influencing a complex disease. Especially for rare variants, we need to analyze multiple variants together to achieve adequate power for detection. In this paper, we propose several parsimonious branching model techniques to assess the joint effect of a group of rare variants in a case-control study. These models implement a data reduction strategy within a likelihood framework and use a weighted score test to assess the statistical significance of the effect of the group of variants on the disease. The primary advantage of the proposed approach is that it performs model-averaging over a substantially smaller set of models supported by the data and thus gains power to detect multi-locus effects. We illustrate these proposed approaches on simulated and real data and study their performance compared to several existing rare variant detection approaches. The primary goal of this paper is to assess if there is any gain in power to detect association by averaging over a number of models instead of selecting the best model. Extensive simulations and real data application demonstrate the advantage the proposed approach in presence of causal variants with opposite directional effects along with a moderate number of null variants in linkage disequilibrium.     

A Statistical Approach for Testing Cross-Phenotype Effects of Rare Variants

Increasing empirical evidence suggests that many genetic variants influence multiple distinct phenotypes. When cross-phenotype effects exist, multivariate association methods that consider pleiotropy are often more powerful than univariate methods that model each phenotype separately. Although several statistical approaches exist for testing cross-phenotype effects for common variants, there is a lack of similar tests for gene-based analysis of rare variants. In order to fill this important gap, we introduce a statistical method for cross-phenotype analysis of rare variants using a nonparametric distance-covariance approach that compares similarity in multivariate phenotypes to similarity in rare-variant genotypes across a gene. The approach can accommodate both binary and continuous phenotypes and further can adjust for covariates. Our approach yields a closed-form test whose significance can be evaluated analytically, thereby improving computational efficiency and permitting application on a genome-wide scale. We use simulated data to demonstrate that our method, which we refer to as the Gene Association with Multiple Traits (GAMuT) test, provides increased power over competing approaches. We also illustrate our approach using exome-chip data from the Genetic Epidemiology Network of Arteriopathy.     

Reconsidering association testing methods using single-variant test statistics as alternatives to pooling tests for sequence data with rare variants

Association tests that pool minor alleles into a measure of burden at a locus have been proposed for case-control studies using sequence data containing rare variants. However, such pooling tests are not robust to the inclusion of neutral and protective variants, which can mask the association signal from risk variants. Early studies proposing pooling tests dismissed methods for locus-wide inference using nonnegative single-variant test statistics based on unrealistic comparisons. However, such methods are robust to the inclusion of neutral and protective variants and therefore may be more useful than previously appreciated. In fact, some recently proposed methods derived within different frameworks are equivalent to performing inference on weighted sums of squared single-variant score statistics. In this study, we compared two existing methods for locus-wide inference using nonnegative single-variant test statistics to two widely cited pooling tests under more realistic conditions. We established analytic results for a simple model with one rare risk and one rare neutral variant, which demonstrated that pooling tests were less powerful than even Bonferroni-corrected single-variant tests in most realistic situations. We also performed simulations using variants with realistic minor allele frequency and linkage disequilibrium spectra, disease models with multiple rare risk variants and extensive neutral variation, and varying rates of missing genotypes. In all scenarios considered, existing methods using nonnegative single-variant test statistics had power comparable to or greater than two widely cited pooling tests. Moreover, in disease models with only rare risk variants, an existing method based on the maximum single-variant Cochran-Armitage trend chi-square statistic in the locus had power comparable to or greater than another existing method closely related to some recently proposed methods. We conclude that efficient locus-wide inference using single-variant test statistics should be reconsidered as a useful framework for devising powerful association tests in sequence data with rare variants.     

Identification of Rare Causal Variants in Sequence-Based Studies: Methods and Applications to VPS13B,a Gene Involved in Cohen Syndrome and Autism

Pinpointing the small number of causal variants among the abundant naturally occurring genetic variation is a difficult challenge, but a crucial one for understanding precise molecular mechanisms of disease and follow-up functional studies. We propose and investigate two complementary statistical approaches for identification of rare causal variants in sequencing studies: a backward elimination procedure based on groupwise association tests, and a hierarchical approach that can integrate sequencing data with diverse functional and evolutionary conservation annotations for individual variants. Using simulations, we show that incorporation of multiple bioinformatic predictors of deleteriousness, such as PolyPhen-2, SIFT and GERP++ scores, can improve the power to discover truly causal variants. As proof of principle, we apply the proposed methods to VPS13B, a gene mutated in the rare neurodevelopmental disorder called Cohen syndrome, and recently reported with recessive variants in autism. We identify a small set of promising candidates for causal variants, including two loss-of-function variants and a rare, homozygous probably-damaging variant that could contribute to autism risk.     

Adaptations to new environments in humans: the role of subtle allele frequency shifts

Humans show tremendous phenotypic diversity across geographically distributed populations, and much of this diversity undoubtedly results from genetic adaptations to different environmental pressures. The availability of genome-wide genetic variation data from densely sampled populations offers unprecedented opportunities for identifying the loci responsible for these adaptations and for elucidating the genetic architecture of human adaptive traits. Several approaches have been used to detect signals of selection in human populations, and these approaches differ in the assumptions they make about the underlying mode of selection. We contrast the results of approaches based on haplotype structure and differentiation of allele frequencies to those from a method for identifying single nucleotide polymorphisms strongly correlated with environmental variables. Although the first group of approaches tends to detect new beneficial alleles that were driven to high frequencies by selection, the environmental correlation approach has power to identify alleles that experienced small shifts in frequency owing to selection. We suggest that the first group of approaches tends to identify only variants with relatively strong phenotypic effects, whereas the environmental correlation methods can detect variants that make smaller contributions to an adaptive trait.     

Family-Based Association Test Using Both Common and Rare Variants and Accounting for Directions of Effects for Sequencing Data

Current family-based association tests for sequencing data were mainly developed for identifying rare variants associated with a complex disease. As the disease can be influenced by the joint effects of common and rare variants, common variants with modest effects may not be identified by the methods focusing on rare variants. Moreover, variants can have risk, neutral, or protective effects. Association tests that can effectively select groups of common and rare variants that are likely to be causal and consider the directions of effects have become important. We developed the Ordered Subset - Variable Threshold - Pedigree Disequilibrium Test (OVPDT), a combination of three algorithms, for association analysis in family sequencing data. The ordered subset algorithm is used to select a subset of common variants based on their relative risks, calculated using only parental mating types. The variable threshold algorithm is used to search for an optimal allele frequency threshold such that rare variants below the threshold are more likely to be causal. The PDT statistics from both rare and common variants selected by the two algorithms are combined as the OVPDT statistic. A permutation procedure is used in OVPDT to calculate the p-value. We used simulations to demonstrate that OVPDT has the correct type I error rates under different scenarios and compared the power of OVPDT with two other family-based association tests. The results suggested that OVPDT can have more power than the other tests if both common and rare variants have effects on the disease in a region.     

Linkage analysis in the next-generation sequencing era

Linkage analysis was developed to detect excess co-segregation of the putative alleles underlying a phenotype with the alleles at a marker locus in family data. Many different variations of this analysis and corresponding study design have been developed to detect this co-segregation. Linkage studies have been shown to have high power to detect loci that have alleles (or variants) with a large effect size, i.e. alleles that make large contributions to the risk of a disease or to the variation of a quantitative trait. However, alleles with a large effect size tend to be rare in the population. In contrast, association studies are designed to have high power to detect common alleles which tend to have a small effect size for most diseases or traits. Although genome-wide association studies have been successful in detecting many new loci with common alleles of small effect for many complex traits, these common variants often do not explain a large proportion of disease risk or variation of the trait. In the past, linkage studies were successful in detecting regions of the genome that were likely to harbor rare variants with large effect for many simple Mendelian diseases and for many complex traits. However, identifying the actual sequence variant(s) responsible for these linkage signals was challenging because of difficulties in sequencing the large regions implicated by each linkage peak. Current 'next-generation' DNA sequencing techniques have made it economically feasible to sequence all exons or the whole genomes of a reasonably large number of individuals. Studies have shown that rare variants are quite common in the general population, and it is now possible to combine these new DNA sequencing methods with linkage studies to identify rare causal variants with a large effect size. A brief review of linkage methods is presented here with examples of their relevance and usefulness for the interpretation of whole-exome and whole-genome sequence data.     

Logistic Bayesian LASSO for identifying association with rare haplotypes and application to age-related macular degeneration

Rare variants have been heralded as key to uncovering "missing heritability" in complex diseases. These variants can now be genotyped using next-generation sequencing technologies; nonetheless, rare haplotypes may also result from combination of common single nucleotide polymorphisms available from genome-wide association studies (GWAS). The National Eye Institute's data on age-related macular degeneration (AMD) is such an example. Studies on AMD had identified potential rare variants; however, due to lack of appropriate statistical tools, effects of individual rare haplotypes were never studied. Here we develop a method for identifying association with rare haplotypes for case-control design. A logistic regression based retrospective likelihood is formulated and is regularized using logistic Bayesian LASSO (LBL). In particular, we penalize the regression coefficients using appropriate priors to weed out unassociated haplotypes, making it possible for the rare associated ones to stand out. We applied LBL to the AMD data and identified common and rare haplotypes in the complement factor H gene, gaining insights into rare variants' contributions to AMD beyond the current literature. This analysis also demonstrates the richness of GWAS data for mapping rare haplotypes-a potential largely unexplored. Additionally, we conducted simulations to investigate the performance of LBL and compare it with Hapassoc. Our results show that LBL is much more powerful in identifying rare associated haplotypes when the false positive rates for both approaches are kept the same.     

A Statistical Approach for Rare-Variant Association Testing in Affected Sibships

Sequencing and exome-chip technologies have motivated development of novel statistical tests to identify rare genetic variation that influences complex diseases. Although many rare-variant association tests exist for case-control or cross-sectional studies, far fewer methods exist for testing association in families. This is unfortunate, because cosegregation of rare variation and disease status in families can amplify association signals for rare variants. Many researchers have begun sequencing (or genotyping via exome chips) familial samples that were either recently collected or previously collected for linkage studies. Because many linkage studies of complex diseases sampled affected sibships, we propose a strategy for association testing of rare variants for use in this study design. The logic behind our approach is that rare susceptibility variants should be found more often on regions shared identical by descent by affected sibling pairs than on regions not shared identical by descent. We propose both burden and variance-component tests of rare variation that are applicable to affected sibships of arbitrary size and that do not require genotype information from unaffected siblings or independent controls. Our approaches are robust to population stratification and produce analytic p values, thereby enabling our approach to scale easily to genome-wide studies of rare variation. We illustrate our methods by using simulated data and exome chip data from sibships ascertained for hypertension collected as part of the Genetic Epidemiology Network of Arteriopathy (GENOA) study.     

Detecting genomic clustering of risk variants from sequence data: cases versus controls

As the ability to measure dense genetic markers approaches the limit of the DNA sequence itself, taking advantage of possible clustering of genetic variants in, and around, a gene would benefit genetic association analyses, and likely provide biological insights. The greatest benefit might be realized when multiple rare variants cluster in a functional region. Several statistical tests have been developed, one of which is based on the popular Kulldorff scan statistic for spatial clustering of disease. We extended another popular spatial clustering method—Tango’s statistic—to genomic sequence data. An advantage of Tango’s method is that it is rapid to compute, and when single test statistic is computed, its distribution is well approximated by a scaled χ ² distribution, making computation of p values very rapid. We compared the Type-I error rates and power of several clustering statistics, as well as the omnibus sequence kernel association test. Although our version of Tango’s statistic, which we call “Kernel Distance” statistic, took approximately half the time to compute than the Kulldorff scan statistic, it had slightly less power than the scan statistic. Our results showed that the Ionita-Laza version of Kulldorff’s scan statistic had the greatest power over a range of clustering scenarios.     

Identification of contemporary selection signatures using composite log likelihood and their associations with marbling score in Korean cattle

Positive selection not only increases beneficial allele frequency but also causes augmentation of allele frequencies of sequence variants in close proximity. Signals for positive selection were detected by the statistical differences in subsequent allele frequencies. To identify selection signatures in Korean cattle, we applied a composite log‐likelihood (CLL)‐based method, which calculates a composite likelihood of the allelic frequencies observed across sliding windows of five adjacent loci and compares the value with the critical statistic estimated by 50 000 permutations. Data for a total of 11 799 nucleotide polymorphisms were used with 71 Korean cattle and 209 foreign beef cattle. As a result, 147 signals were identified for Korean cattle based on CLL estimates (P < 0.01). The signals might be candidate genetic factors for meat quality by which the Korean cattle have been selected. Further genetic association analysis with 41 intragenic variants in the selection signatures with the greatest CLL for each chromosome revealed that marbling score was associated with five variants. Intensive association studies with all the selection signatures identified in this study are required to exclude signals associated with other phenotypes or signals falsely detected and thus to identify genetic markers for meat quality.     

Rare-variant association testing for sequencing data with the sequence kernel association test

Sequencing studies are increasingly being conducted to identify rare variants associated with complex traits. The limited power of classical single-marker association analysis for rare variants poses a central challenge in such studies. We propose the sequence kernel association test (SKAT), a supervised, flexible, computationally efficient regression method to test for association between genetic variants (common and rare) in a region and a continuous or dichotomous trait while easily adjusting for covariates. As a score-based variance-component test, SKAT can quickly calculate p values analytically by fitting the null model containing only the covariates, and so can easily be applied to genome-wide data. Using SKAT to analyze a genome-wide sequencing study of 1000 individuals, by segmenting the whole genome into 30 kb regions, requires only 7 hr on a laptop. Through analysis of simulated data across a wide range of practical scenarios and triglyceride data from the Dallas Heart Study, we show that SKAT can substantially outperform several alternative rare-variant association tests. We also provide analytic power and sample-size calculations to help design candidate-gene, whole-exome, and whole-genome sequence association studies.     

Studying gene and gene-environment effects of uncommon and common variants on continuous traits: a marker-set approach using gene-trait similarity regression

Genomic association analyses of complex traits demand statistical tools that are capable of detecting small effects of common and rare variants and modeling complex interaction effects and yet are computationally feasible. In this work, we introduce a similarity-based regression method for assessing the main genetic and interaction effects of a group of markers on quantitative traits. The method uses genetic similarity to aggregate information from multiple polymorphic sites and integrates adaptive weights that depend on allele frequencies to accomodate common and uncommon variants. Collapsing information at the similarity level instead of the genotype level avoids canceling signals that have the opposite etiological effects and is applicable to any class of genetic variants without the need for dichotomizing the allele types. To assess gene-trait associations, we regress trait similarities for pairs of unrelated individuals on their genetic similarities and assess association by using a score test whose limiting distribution is derived in this work. The proposed regression framework allows for covariates, has the capacity to model both main and interaction effects, can be applied to a mixture of different polymorphism types, and is computationally efficient. These features make it an ideal tool for evaluating associations between phenotype and marker sets defined by linkage disequilibrium (LD) blocks, genes, or pathways in whole-genome analysis.     

Methods for detecting associations with rare variants for common diseases: application to analysis of sequence data

Although whole-genome association studies using tagSNPs are a powerful approach for detecting common variants, they are underpowered for detecting associations with rare variants. Recent studies have demonstrated that common diseases can be due to functional variants with a wide spectrum of allele frequencies, ranging from rare to common. An effective way to identify rare variants is through direct sequencing. The development of cost-effective sequencing technologies enables association studies to use sequence data from candidate genes and, in the future, from the entire genome. Although methods used for analysis of common variants are applicable to sequence data, their performance might not be optimal. In this study, it is shown that the collapsing method, which involves collapsing genotypes across variants and applying a univariate test, is powerful for analyzing rare variants, whereas multivariate analysis is robust against inclusion of noncausal variants. Both methods are superior to analyzing each variant individually with univariate tests. In order to unify the advantages of both collapsing and multiple-marker tests, we developed the Combined Multivariate and Collapsing (CMC) method and demonstrated that the CMC method is both powerful and robust. The CMC method can be applied to either candidate-gene or whole-genome sequence data.     

Rare and Low Frequency Variant Stratification in the UK Population: Description and Impact on Association Tests

Although variations in allele frequencies at common SNPs have been extensively studied in different populations, little is known about the stratification of rare variants and its impact on association tests. In this paper, we used Affymetrix 500K genotype data from the WTCCC to investigate if variants in three different frequency categories (below 1%, between 1 and 5%, above 5%) show different stratification patterns in the UK population. We found that these patterns are indeed different. The top principal component extracted from the rare variant category shows poor correlations with any principal component or combination of principal components from the low frequency or common variant categories. These results could suggest that a suitable solution to avoid false positive association due to population stratification would involve adjusting for the respective PCs when testing for variants in different allele frequency categories. However, we found this was not the case both on type 2 diabetes data and on simulated data. Indeed, adjusting rare variant association tests on PCs derived from rare variants does no better to correct for population stratification than adjusting on PCs derived from more common variants. Mixed models perform slightly better for low frequency variants than PC based adjustments but less well for the rarest variants. These results call for the need of new methodological developments specifically devoted to address rare variant stratification issues in association tests.