Conditional Random Fields for Fast, Large-Scale Genome-Wide Association Studies

Understanding the role of genetic variation in human diseases remains an important problem to be solved in genomics. An important component of such variation consist of variations at single sites in DNA, or single nucleotide polymorphisms (SNPs). Typically, the problem of associating particular SNPs to phenotypes has been confounded by hidden factors such as the presence of population structure, family structure or cryptic relatedness in the sample of individuals being analyzed. Such confounding factors lead to a large number of spurious associations and missed associations. Various statistical methods have been proposed to account for such confounding factors such as linear mixed-effect models (LMMs) or methods that adjust data based on a principal components analysis (PCA), but these methods either suffer from low power or cease to be tractable for larger numbers of individuals in the sample. Here we present a statistical model for conducting genome-wide association studies (GWAS) that accounts for such confounding factors. Our method scales in runtime quadratic in the number of individuals being studied with only a modest loss in statistical power as compared to LMM-based and PCA-based methods when testing on synthetic data that was generated from a generalized LMM. Applying our method to both real and synthetic human genotype/phenotype data, we demonstrate the ability of our model to correct for confounding factors while requiring significantly less runtime relative to LMMs. We have implemented methods for fitting these models, which are available at http://www.microsoft.com/science.     

Exploiting Population Samples to Enhance Genome-Wide Association Studies of Disease

It is widely acknowledged that genome-wide association studies (GWAS) of complex human disease fail to explain a large portion of heritability, primarily due to lack of statistical power—a problem that is exacerbated when seeking detection of interactions of multiple genomic loci. An untapped source of information that is already widely available, and that is expected to grow in coming years, is population samples. Such samples contain genetic marker data for additional individuals, but not their relevant phenotypes. In this article we develop a highly efficient testing framework based on a constrained maximum-likelihood estimate in a case–control–population setting. We leverage the available population data and optional modeling assumptions, such as Hardy–Weinberg equilibrium (HWE) in the population and linkage equilibrium (LE) between distal loci, to substantially improve power of association and interaction tests. We demonstrate, via simulation and application to actual GWAS data sets, that our approach is substantially more powerful and robust than standard testing approaches that ignore or make naive use of the population sample. We report several novel and credible pairwise interactions, in bipolar disorder, coronary artery disease, Crohn’s disease, and rheumatoid arthritis.     

SNP2GO: Functional Analysis of Genome-Wide Association Studies

Genome-wide association studies (GWAS) are designed to identify the portion of single-nucleotide polymorphisms (SNPs) in genome sequences associated with a complex trait. Strategies based on the gene list enrichment concept are currently applied for the functional analysis of GWAS, according to which a significant overrepresentation of candidate genes associated with a biological pathway is used as a proxy to infer overrepresentation of candidate SNPs in the pathway. Here we show that such inference is not always valid and introduce the program SNP2GO, which implements a new method to properly test for the overrepresentation of candidate SNPs in biological pathways.     

GStream: Improving SNP and CNV Coverage on Genome-Wide Association Studies

We present GStream, a method that combines genome-wide SNP and CNV genotyping in the Illumina microarray platform with unprecedented accuracy. This new method outperforms previous well-established SNP genotyping software. More importantly, the CNV calling algorithm of GStream dramatically improves the results obtained by previous state-of-the-art methods and yields an accuracy that is close to that obtained by purely CNV-oriented technologies like Comparative Genomic Hybridization (CGH). We demonstrate the superior performance of GStream using microarray data generated from HapMap samples. Using the reference CNV calls generated by the 1000 Genomes Project (1KGP) and well-known studies on whole genome CNV characterization based either on CGH or genotyping microarray technologies, we show that GStream can increase the number of reliably detected variants up to 25% compared to previously developed methods. Furthermore, the increased genome coverage provided by GStream allows the discovery of CNVs in close linkage disequilibrium with SNPs, previously associated with disease risk in published Genome-Wide Association Studies (GWAS). These results could provide important insights into the biological mechanism underlying the detected disease risk association. With GStream, large-scale GWAS will not only benefit from the combined genotyping of SNPs and CNVs at an unprecedented accuracy, but will also take advantage of the computational efficiency of the method.     

Iterative Usage of Fixed and Random Effect Models for Powerful and Efficient Genome-Wide Association Studies

False positives in a Genome-Wide Association Study (GWAS) can be effectively controlled by a fixed effect and random effect Mixed Linear Model (MLM) that incorporates population structure and kinship among individuals to adjust association tests on markers; however, the adjustment also compromises true positives. The modified MLM method, Multiple Loci Linear Mixed Model (MLMM), incorporates multiple markers simultaneously as covariates in a stepwise MLM to partially remove the confounding between testing markers and kinship. To completely eliminate the confounding, we divided MLMM into two parts: Fixed Effect Model (FEM) and a Random Effect Model (REM) and use them iteratively. FEM contains testing markers, one at a time, and multiple associated markers as covariates to control false positives. To avoid model over-fitting problem in FEM, the associated markers are estimated in REM by using them to define kinship. The P values of testing markers and the associated markers are unified at each iteration. We named the new method as Fixed and random model Circulating Probability Unification (FarmCPU). Both real and simulated data analyses demonstrated that FarmCPU improves statistical power compared to current methods. Additional benefits include an efficient computing time that is linear to both number of individuals and number of markers. Now, a dataset with half million individuals and half million markers can be analyzed within three days.     

Commercially Available Outbred Mice for Genome-Wide Association Studies

Genome-wide association studies using commercially available outbred mice can detect genes involved in phenotypes of biomedical interest. Useful populations need high-frequency alleles to ensure high power to detect quantitative trait loci (QTLs), low linkage disequilibrium between markers to obtain accurate mapping resolution, and an absence of population structure to prevent false positive associations. We surveyed 66 colonies for inbreeding, genetic diversity, and linkage disequilibrium, and we demonstrate that some have haplotype blocks of less than 100 Kb, enabling gene-level mapping resolution. The same alleles contribute to variation in different colonies, so that when mapping progress stalls in one, another can be used in its stead. Colonies are genetically diverse: 45% of the total genetic variation is attributable to differences between colonies. However, quantitative differences in allele frequencies, rather than the existence of private alleles, are responsible for these population differences. The colonies derive from a limited pool of ancestral haplotypes resembling those found in inbred strains: over 95% of sequence variants segregating in outbred populations are found in inbred strains. Consequently it is possible to impute the sequence of any mouse from a dense SNP map combined with inbred strain sequence data, which opens up the possibility of cataloguing and testing all variants for association, a situation that has so far eluded studies in completely outbred populations. We demonstrate the colonies'' potential by identifying a deletion in the promoter of H2-Ea as the molecular change that strongly contributes to setting the ratio of CD4+ and CD8+ lymphocytes.     

Fidelity of SNP Array Genotyping Using Epstein Barr Virus-Transformed B-Lymphocyte Cell Lines: Implications for Genome-Wide Association Studies

Background

As availability of primary cells can be limited for genetic studies of human disease, lymphoblastoid cell lines (LCL) are common sources of genomic DNA. LCL are created in a transformation process that entails in vitro infection of human B-lymphocytes with the Epstein-Barr Virus (EBV).

Methodology/Principal Findings

To test for genotypic errors potentially induced by the Epstein-Barr Virus transformation process, we compared single nucleotide polymorphism (SNP) genotype calls in peripheral blood mononuclear cells (PBMC) and LCL from the same individuals. The average mismatch rate across 19 comparisons was 0.12% for SNPs with a population call rate of at least 95%, and 0.03% at SNPs with a call rate of at least 99%. Mismatch rates were not correlated across genotype subarrays run on all sample pairs.

Conclusions/Significance

Genotypic discrepancies found in PBMC and LCL pairs were not significantly different than control pairs, and were not correlated across subarrays. These results suggest that mismatch rates are minimal with stringent quality control, and that most genotypic discrepancies are due to technical artifacts rather than the EBV transformation process. Thus, LCL likely constitute a reliable DNA source for host genotype analysis.     

Chapter 11: Genome-Wide Association Studies

Genome-wide association studies (GWAS) have evolved over the last ten years into a powerful tool for investigating the genetic architecture of human disease. In this work, we review the key concepts underlying GWAS, including the architecture of common diseases, the structure of common human genetic variation, technologies for capturing genetic information, study designs, and the statistical methods used for data analysis. We also look forward to the future beyond GWAS.

What to Learn in This Chapter

• Basic genetic concepts that drive genome-wide association studies
• Genotyping technologies and common study designs
• Statistical concepts for GWAS analysis
• Replication, interpretation, and follow-up of association results

This article is part of the “Translational Bioinformatics” collection for PLOS Computational Biology.

     

Statistical Power of Model Selection Strategies for Genome-Wide Association Studies

Genome-wide association studies (GWAS) aim to identify genetic variants related to diseases by examining the associations between phenotypes and hundreds of thousands of genotyped markers. Because many genes are potentially involved in common diseases and a large number of markers are analyzed, it is crucial to devise an effective strategy to identify truly associated variants that have individual and/or interactive effects, while controlling false positives at the desired level. Although a number of model selection methods have been proposed in the literature, including marginal search, exhaustive search, and forward search, their relative performance has only been evaluated through limited simulations due to the lack of an analytical approach to calculating the power of these methods. This article develops a novel statistical approach for power calculation, derives accurate formulas for the power of different model selection strategies, and then uses the formulas to evaluate and compare these strategies in genetic model spaces. In contrast to previous studies, our theoretical framework allows for random genotypes, correlations among test statistics, and a false-positive control based on GWAS practice. After the accuracy of our analytical results is validated through simulations, they are utilized to systematically evaluate and compare the performance of these strategies in a wide class of genetic models. For a specific genetic model, our results clearly reveal how different factors, such as effect size, allele frequency, and interaction, jointly affect the statistical power of each strategy. An example is provided for the application of our approach to empirical research. The statistical approach used in our derivations is general and can be employed to address the model selection problems in other random predictor settings. We have developed an R package markerSearchPower to implement our formulas, which can be downloaded from the Comprehensive R Archive Network (CRAN) or http://bioinformatics.med.yale.edu/group/.     

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.     

Exploiting Protein-Protein Interaction Networks for Genome-Wide Disease-Gene Prioritization

Complex genetic disorders often involve products of multiple genes acting cooperatively. Hence, the pathophenotype is the outcome of the perturbations in the underlying pathways, where gene products cooperate through various mechanisms such as protein-protein interactions. Pinpointing the decisive elements of such disease pathways is still challenging. Over the last years, computational approaches exploiting interaction network topology have been successfully applied to prioritize individual genes involved in diseases. Although linkage intervals provide a list of disease-gene candidates, recent genome-wide studies demonstrate that genes not associated with any known linkage interval may also contribute to the disease phenotype. Network based prioritization methods help highlighting such associations. Still, there is a need for robust methods that capture the interplay among disease-associated genes mediated by the topology of the network. Here, we propose a genome-wide network-based prioritization framework named GUILD. This framework implements four network-based disease-gene prioritization algorithms. We analyze the performance of these algorithms in dozens of disease phenotypes. The algorithms in GUILD are compared to state-of-the-art network topology based algorithms for prioritization of genes. As a proof of principle, we investigate top-ranking genes in Alzheimer''s disease (AD), diabetes and AIDS using disease-gene associations from various sources. We show that GUILD is able to significantly highlight disease-gene associations that are not used a priori. Our findings suggest that GUILD helps to identify genes implicated in the pathology of human disorders independent of the loci associated with the disorders.     

Genome-Wide Association Studies of the Human Gut Microbiota

The bacterial composition of the human fecal microbiome is influenced by many lifestyle factors, notably diet. It is less clear, however, what role host genetics plays in dictating the composition of bacteria living in the gut. In this study, we examined the association of ~200K host genotypes with the relative abundance of fecal bacterial taxa in a founder population, the Hutterites, during two seasons (n = 91 summer, n = 93 winter, n = 57 individuals collected in both). These individuals live and eat communally, minimizing variation due to environmental exposures, including diet, which could potentially mask small genetic effects. Using a GWAS approach that takes into account the relatedness between subjects, we identified at least 8 bacterial taxa whose abundances were associated with single nucleotide polymorphisms in the host genome in each season (at genome-wide FDR of 20%). For example, we identified an association between a taxon known to affect obesity (genus Akkermansia) and a variant near PLD1, a gene previously associated with body mass index. Moreover, we replicate a previously reported association from a quantitative trait locus (QTL) mapping study of fecal microbiome abundance in mice (genus Lactococcus, rs3747113, P = 3.13 x 10⁻⁷). Finally, based on the significance distribution of the associated microbiome QTLs in our study with respect to chromatin accessibility profiles, we identified tissues in which host genetic variation may be acting to influence bacterial abundance in the gut.     

Use of Genome-Wide Association Studies for Cancer Research and Drug Repositioning

Although genome-wide association studies have identified many risk loci associated with colorectal cancer, the molecular basis of these associations are still unclear. We aimed to infer biological insights and highlight candidate genes of interest within GWAS risk loci. We used an in silico pipeline based on functional annotation, quantitative trait loci mapping of cis-acting gene, PubMed text-mining, protein-protein interaction studies, genetic overlaps with cancer somatic mutations and knockout mouse phenotypes, and functional enrichment analysis to prioritize the candidate genes at the colorectal cancer risk loci. Based on these analyses, we observed that these genes were the targets of approved therapies for colorectal cancer, and suggested that drugs approved for other indications may be repurposed for the treatment of colorectal cancer. This study highlights the use of publicly available data as a cost effective solution to derive biological insights, and provides an empirical evidence that the molecular basis of colorectal cancer can provide important leads for the discovery of new drugs.     

Genome-Wide Association Studies for Growth and Meat Production Traits in Sheep

Background

Growth and meat production traits are significant economic traits in sheep. The aim of the study is to identify candidate genes affecting growth and meat production traits at genome level with high throughput single nucleotide polymorphisms (SNP) genotyping technologies.

Methodology and Results

Using Illumina OvineSNP50 BeadChip, we performed a GWA study in 329 purebred sheep for 11 growth and meat production traits (birth weight, weaning weight, 6-month weight, eye muscle area, fat thickness, pre-weaning gain, post-weaning gain, daily weight gain, height at withers, chest girth, and shin circumference). After quality control, 319 sheep and 48,198 SNPs were analyzed by TASSEL program in a mixed linear model (MLM). 36 significant SNPs were identified for 7 traits, and 10 of them reached genome-wise significance level for post-weaning gain. Gene annotation was implemented with the latest sheep genome Ovis_aries_v3.1 (released October 2012). More than one-third SNPs (14 out of 36) were located within ovine genes, others were located close to ovine genes (878bp-398,165bp apart). The strongest new finding is 5 genes were thought to be the most crucial candidate genes associated with post-weaning gain: s58995.1 was located within the ovine genes MEF2B and RFXANK, OAR3_84073899.1, OAR3_115712045.1 and OAR9_91721507.1 were located within CAMKMT, TRHDE, and RIPK2 respectively. GRM1, POL, MBD5, UBR2, RPL7 and SMC2 were thought to be the important candidate genes affecting post-weaning gain too. Additionally, 25 genes at chromosome-wise significance level were also forecasted to be the promising genes that influencing sheep growth and meat production traits.

Conclusions

The results will contribute to the similar studies and facilitate the potential utilization of genes involved in growth and meat production traits in sheep in future.     

Adjusting for Heritable Covariates Can Bias Effect Estimates in Genome-Wide Association Studies

In recent years, a number of large-scale genome-wide association studies have been published for human traits adjusted for other correlated traits with a genetic basis. In most studies, the motivation for such an adjustment is to discover genetic variants associated with the primary outcome independently of the correlated trait. In this report, we contend that this objective is fulfilled when the tested variants have no effect on the covariate or when the correlation between the covariate and the outcome is fully explained by a direct effect of the covariate on the outcome. For all other scenarios, an unintended bias is introduced with respect to the primary outcome as a result of the adjustment, and this bias might lead to false positives. Here, we illustrate this point by providing examples from published genome-wide association studies, including large meta-analysis of waist-to-hip ratio and waist circumference adjusted for body mass index (BMI), where genetic effects might be biased as a result of adjustment for body mass index. Using both theory and simulations, we explore this phenomenon in detail and discuss the ramifications for future genome-wide association studies of correlated traits and diseases.     

Principal Component Analysis Characterizes Shared Pathogenetics from Genome-Wide Association Studies

Genome-wide association studies (GWASs) have recently revealed many genetic associations that are shared between different diseases. We propose a method, disPCA, for genome-wide characterization of shared and distinct risk factors between and within disease classes. It flips the conventional GWAS paradigm by analyzing the diseases themselves, across GWAS datasets, to explore their “shared pathogenetics”. The method applies principal component analysis (PCA) to gene-level significance scores across all genes and across GWASs, thereby revealing shared pathogenetics between diseases in an unsupervised fashion. Importantly, it adjusts for potential sources of heterogeneity present between GWAS which can confound investigation of shared disease etiology. We applied disPCA to 31 GWASs, including autoimmune diseases, cancers, psychiatric disorders, and neurological disorders. The leading principal components separate these disease classes, as well as inflammatory bowel diseases from other autoimmune diseases. Generally, distinct diseases from the same class tend to be less separated, which is in line with their increased shared etiology. Enrichment analysis of genes contributing to leading principal components revealed pathways that are implicated in the immune system, while also pointing to pathways that have yet to be explored before in this context. Our results point to the potential of disPCA in going beyond epidemiological findings of the co-occurrence of distinct diseases, to highlighting novel genes and pathways that unsupervised learning suggest to be key players in the variability across diseases.     

Replication Study in a Japanese Population to Evaluate the Association between 10 SNP Loci,Identified in European Genome-Wide Association Studies,and Type 2 Diabetes

AimWe performed a replication study in a Japanese population to evaluate the association between type 2 diabetes and 7 susceptibility loci originally identified by European genome-wide association study (GWAS) in 2012: ZMIZ1, KLHDC5, TLE1, ANKRD55, CILP2, MC4R, and BCAR1. We also examined the association of 3 additional loci: CCND2 and GIPR, identified in sex-differentiated analyses, and LAMA1, which was shown to be associated with non-obese European type 2 diabetes.MethodsWe genotyped 6,972 Japanese participants (4,280 type 2 diabetes patients and 2,692 controls) for each of the 10 single nucleotide polymorphisms (SNPs): rs12571751 in ZMIZ1, rs10842994 near KLHDC5, rs2796441 near TLE1, rs459193 near ANKRD55, rs10401969 in CILP2, rs12970134 near MC4R, rs7202877 near BCAR1, rs11063069 near CCND2, rs8108269 near GIPR, and rs8090011 in LAMA1 using a multiplex polymerase chain reaction invader assay. The association of each SNP locus with the disease was evaluated using a logistic regression analysis.ResultsAll SNPs examined in this study had the same direction of effect (odds ratio > 1.0, p = 9.77 × 10^-4, binomial test), as in the original reports. Among them, rs12571751 in ZMIZ1 was significantly associated with type 2 diabetes [p = 0.0041, odds ratio = 1.123, 95% confidence interval 1.037–1.215, adjusted for sex, age and body mass index (BMI)], but we did not observe significant association of the remaining 9 SNP loci with type 2 diabetes in the present Japanese population (p ≥ 0.005). A genetic risk score, constructed from the sum of risk alleles for the 7 SNP loci identified by un-stratified analyses in the European GWAS meta-analysis were associated with type 2 diabetes in the present Japanese population (p = 2.3 × 10^-4, adjusted for sex, age and BMI).ConclusionsZMIZ1 locus has a significant effect on conferring susceptibility to type 2 diabetes also in the Japanese population.     

Analysis of Genome-Wide Association Studies with Multiple Outcomes Using Penalization

Genome-wide association studies have been extensively conducted, searching for markers for biologically meaningful outcomes and phenotypes. Penalization methods have been adopted in the analysis of the joint effects of a large number of SNPs (single nucleotide polymorphisms) and marker identification. This study is partly motivated by the analysis of heterogeneous stock mice dataset, in which multiple correlated phenotypes and a large number of SNPs are available. Existing penalization methods designed to analyze a single response variable cannot accommodate the correlation among multiple response variables. With multiple response variables sharing the same set of markers, joint modeling is first employed to accommodate the correlation. The group Lasso approach is adopted to select markers associated with all the outcome variables. An efficient computational algorithm is developed. Simulation study and analysis of the heterogeneous stock mice dataset show that the proposed method can outperform existing penalization methods.