Search

Genome and exome sequencing in large cohorts enables characterization of the role of rare variation in complex diseases. Success in this endeavor, however, requires investigators to test a diverse array of genetic hypotheses which differ in the number, frequency and effect sizes of underlying causal variants. In this study, we evaluated the power of gene-based association methods to interrogate such hypotheses, and examined the implications for study design. We developed a flexible simulation approach, using 1000 Genomes data, to (a) generate sequence variation at human genes in up to 10K case-control samples, and (b) quantify the statistical power of a panel of widely used gene-based association tests under a variety of allelic architectures, locus effect sizes, and significance thresholds. For loci explaining ~1% of phenotypic variance underlying a common dichotomous trait, we find that all methods have low absolute power to achieve exome-wide significance (~5-20% power at α = 2.5 × 10(-6)) in 3K individuals; even in 10K samples, power is modest (~60%). The combined application of multiple methods increases sensitivity, but does so at the expense of a higher false positive rate. MiST, SKAT-O, and KBAC have the highest individual mean power across simulated datasets, but we observe wide architecture-dependent variability in the individual loci detected by each test, suggesting that inferences about disease architecture from analysis of sequencing studies can differ depending on which methods are used. Our results imply that tens of thousands of individuals, extensive functional annotation, or highly targeted hypothesis testing will be required to confidently detect or exclude rare variant signals at complex disease loci.

Genome wide association studies (GWAS) for fasting glucose (FG) and insulin (FI) have identified common variant signals which explain 4.8% and 1.2% of trait variance, respectively. It is hypothesized that low-frequency and rare variants could contribute substantially to unexplained genetic variance. To test this, we analyzed exome-array data from up to 33,231 non-diabetic individuals of European ancestry. We found exome-wide significant (P

Common genetic variants contribute to the observed variation in breast cancer risk for BRCA2 mutation carriers; those known to date have all been found through population-based genome-wide association studies (GWAS). To comprehensively identify breast cancer risk modifying loci for BRCA2 mutation carriers, we conducted a deep replication of an ongoing GWAS discovery study. Using the ranked P-values of the breast cancer associations with the imputed genotype of 1.4 M SNPs, 19,029 SNPs were selected and designed for inclusion on a custom Illumina array that included a total of 211,155 SNPs as part of a multi-consortial project. DNA samples from 3,881 breast cancer affected and 4,330 unaffected BRCA2 mutation carriers from 47 studies belonging to the Consortium of Investigators of Modifiers of BRCA1/2 were genotyped and available for analysis. We replicated previously reported breast cancer susceptibility alleles in these BRCA2 mutation carriers and for several regions (including FGFR2, MAP3K1, CDKN2A/B, and PTHLH) identified SNPs that have stronger evidence of association than those previously published. We also identified a novel susceptibility allele at 6p24 that was inversely associated with risk in BRCA2 mutation carriers (rs9348512; per allele HR = 0.85, 95% CI 0.80-0.90, P = 3.9 × 10(-8)). This SNP was not associated with breast cancer risk either in the general population or in BRCA1 mutation carriers. The locus lies within a region containing TFAP2A, which encodes a transcriptional activation protein that interacts with several tumor suppressor genes. This report identifies the first breast cancer risk locus specific to a BRCA2 mutation background. This comprehensive update of novel and previously reported breast cancer susceptibility loci contributes to the establishment of a panel of SNPs that modify breast cancer risk in BRCA2 mutation carriers. This panel may have clinical utility for women with BRCA2 mutations weighing options for medical prevention of breast cancer.

Circulating levels of adiponectin, a hormone produced predominantly by adipocytes, are highly heritable and are inversely associated with type 2 diabetes mellitus (T2D) and other metabolic traits. We conducted a meta-analysis of genome-wide association studies in 39,883 individuals of European ancestry to identify genes associated with metabolic disease. We identified 8 novel loci associated with adiponectin levels and confirmed 2 previously reported loci (P = 4.5×10(-8)-1.2×10(-43)). Using a novel method to combine data across ethnicities (N = 4,232 African Americans, N = 1,776 Asians, and N = 29,347 Europeans), we identified two additional novel loci. Expression analyses of 436 human adipocyte samples revealed that mRNA levels of 18 genes at candidate regions were associated with adiponectin concentrations after accounting for multiple testing (p

Genome-wide association studies have identified hundreds of loci for type 2 diabetes, coronary artery disease and myocardial infarction, as well as for related traits such as body mass index, glucose and insulin levels, lipid levels, and blood pressure. These studies also have pointed to thousands of loci with promising but not yet compelling association evidence. To establish association at additional loci and to characterize the genome-wide significant loci by fine-mapping, we designed the "Metabochip," a custom genotyping array that assays nearly 200,000 SNP markers. Here, we describe the Metabochip and its component SNP sets, evaluate its performance in capturing variation across the allele-frequency spectrum, describe solutions to methodological challenges commonly encountered in its analysis, and evaluate its performance as a platform for genotype imputation. The metabochip achieves dramatic cost efficiencies compared to designing single-trait follow-up reagents, and provides the opportunity to compare results across a range of related traits. The metabochip and similar custom genotyping arrays offer a powerful and cost-effective approach to follow-up large-scale genotyping and sequencing studies and advance our understanding of the genetic basis of complex human diseases and traits.

Genetic case-control association studies often include data on clinical covariates, such as body mass index (BMI), smoking status, or age, that may modify the underlying genetic risk of case or control samples. For example, in type 2 diabetes, odds ratios for established variants estimated from low-BMI cases are larger than those estimated from high-BMI cases. An unanswered question is how to use this information to maximize statistical power in case-control studies that ascertain individuals on the basis of phenotype (case-control ascertainment) or phenotype and clinical covariates (case-control-covariate ascertainment). While current approaches improve power in studies with random ascertainment, they often lose power under case-control ascertainment and fail to capture available power increases under case-control-covariate ascertainment. We show that an informed conditioning approach, based on the liability threshold model with parameters informed by external epidemiological information, fully accounts for disease prevalence and non-random ascertainment of phenotype as well as covariates and provides a substantial increase in power while maintaining a properly controlled false-positive rate. Our method outperforms standard case-control association tests with or without covariates, tests of gene x covariate interaction, and previously proposed tests for dealing with covariates in ascertained data, with especially large improvements in the case of case-control-covariate ascertainment. We investigate empirical case-control studies of type 2 diabetes, prostate cancer, lung cancer, breast cancer, rheumatoid arthritis, age-related macular degeneration, and end-stage kidney disease over a total of 89,726 samples. In these datasets, informed conditioning outperforms logistic regression for 115 of the 157 known associated variants investigated (P-value = 1 × 10(-9)). The improvement varied across diseases with a 16% median increase in χ(2) test statistics and a commensurate increase in power. This suggests that applying our method to existing and future association studies of these diseases may identify novel disease loci.

Weight-loss interventions generally improve lipid profiles and reduce cardiovascular disease risk, but effects are variable and may depend on genetic factors. We performed a genetic association analysis of data from 2,993 participants in the Diabetes Prevention Program to test the hypotheses that a genetic risk score (GRS) based on deleterious alleles at 32 lipid-associated single-nucleotide polymorphisms modifies the effects of lifestyle and/or metformin interventions on lipid levels and nuclear magnetic resonance (NMR) lipoprotein subfraction size and number. Twenty-three loci previously associated with fasting LDL-C, HDL-C, or triglycerides replicated (P = 0.04-1 × 10(-17)). Except for total HDL particles (r = -0.03, P = 0.26), all components of the lipid profile correlated with the GRS (partial |r| = 0.07-0.17, P = 5 × 10(-5)-1 10(-19)). The GRS was associated with higher baseline-adjusted 1-year LDL cholesterol levels (β = +0.87, SEE ± 0.22 mg/dl/allele, P = 8 × 10(-5), P(interaction) = 0.02) in the lifestyle intervention group, but not in the placebo (β = +0.20, SEE ± 0.22 mg/dl/allele, P = 0.35) or metformin (β = -0.03, SEE ± 0.22 mg/dl/allele, P = 0.90; P(interaction) = 0.64) groups. Similarly, a higher GRS predicted a greater number of baseline-adjusted small LDL particles at 1 year in the lifestyle intervention arm (β = +0.30, SEE ± 0.012 ln nmol/L/allele, P = 0.01, P(interaction) = 0.01) but not in the placebo (β = -0.002, SEE ± 0.008 ln nmol/L/allele, P = 0.74) or metformin (β = +0.013, SEE ± 0.008 nmol/L/allele, P = 0.12; P(interaction) = 0.24) groups. Our findings suggest that a high genetic burden confers an adverse lipid profile and predicts attenuated response in LDL-C levels and small LDL particle number to dietary and physical activity interventions aimed at weight loss.

Technological advances make it possible to use high-throughput sequencing as a primary discovery tool of medical genetics, specifically for assaying rare variation. Still this approach faces the analytic challenge that the influence of very rare variants can only be evaluated effectively as a group. A further complication is that any given rare variant could have no effect, could increase risk, or could be protective. We propose here the C-alpha test statistic as a novel approach for testing for the presence of this mixture of effects across a set of rare variants. Unlike existing burden tests, C-alpha, by testing the variance rather than the mean, maintains consistent power when the target set contains both risk and protective variants. Through simulations and analysis of case/control data, we demonstrate good power relative to existing methods that assess the burden of rare variants in individuals.

Investigators have linked rare copy number variation (CNVs) to neuropsychiatric diseases, such as schizophrenia. One hypothesis is that CNV events cause disease by affecting genes with specific brain functions. Under these circumstances, we expect that CNV events in cases should impact brain-function genes more frequently than those events in controls. Previous publications have applied "pathway" analyses to genes within neuropsychiatric case CNVs to show enrichment for brain-functions. While such analyses have been suggestive, they often have not rigorously compared the rates of CNVs impacting genes with brain function in cases to controls, and therefore do not address important confounders such as the large size of brain genes and overall differences in rates and sizes of CNVs. To demonstrate the potential impact of confounders, we genotyped rare CNV events in 2,415 unaffected controls with Affymetrix 6.0; we then applied standard pathway analyses using four sets of brain-function genes and observed an apparently highly significant enrichment for each set. The enrichment is simply driven by the large size of brain-function genes. Instead, we propose a case-control statistical test, cnv-enrichment-test, to compare the rate of CNVs impacting specific gene sets in cases versus controls. With simulations, we demonstrate that cnv-enrichment-test is robust to case-control differences in CNV size, CNV rate, and systematic differences in gene size. Finally, we apply cnv-enrichment-test to rare CNV events published by the International Schizophrenia Consortium (ISC). This approach reveals nominal evidence of case-association in neuronal-activity and the learning gene sets, but not the other two examined gene sets. The neuronal-activity genes have been associated in a separate set of schizophrenia cases and controls; however, testing in independent samples is necessary to definitively confirm this association. Our method is implemented in the PLINK software package.

Mitochondrial dysfunction has been observed in skeletal muscle of people with diabetes and insulin-resistant individuals. Furthermore, inherited mutations in mitochondrial DNA can cause a rare form of diabetes. However, it is unclear whether mitochondrial dysfunction is a primary cause of the common form of diabetes. To date, common genetic variants robustly associated with type 2 diabetes (T2D) are not known to affect mitochondrial function. One possibility is that multiple mitochondrial genes contain modest genetic effects that collectively influence T2D risk. To test this hypothesis we developed a method named Meta-Analysis Gene-set Enrichment of variaNT Associations (MAGENTA; http://www.broadinstitute.org/mpg/magenta). MAGENTA, in analogy to Gene Set Enrichment Analysis, tests whether sets of functionally related genes are enriched for associations with a polygenic disease or trait. MAGENTA was specifically designed to exploit the statistical power of large genome-wide association (GWA) study meta-analyses whose individual genotypes are not available. This is achieved by combining variant association p-values into gene scores and then correcting for confounders, such as gene size, variant number, and linkage disequilibrium properties. Using simulations, we determined the range of parameters for which MAGENTA can detect associations likely missed by single-marker analysis. We verified MAGENTA's performance on empirical data by identifying known relevant pathways in lipid and lipoprotein GWA meta-analyses. We then tested our mitochondrial hypothesis by applying MAGENTA to three gene sets: nuclear regulators of mitochondrial genes, oxidative phosphorylation genes, and approximately 1,000 nuclear-encoded mitochondrial genes. The analysis was performed using the most recent T2D GWA meta-analysis of 47,117 people and meta-analyses of seven diabetes-related glycemic traits (up to 46,186 non-diabetic individuals). This well-powered analysis found no significant enrichment of associations to T2D or any of the glycemic traits in any of the gene sets tested. These results suggest that common variants affecting nuclear-encoded mitochondrial genes have at most a small genetic contribution to T2D susceptibility.

It has been recently hypothesized that many of the signals detected in genome-wide association studies (GWAS) to T2D and other diseases, despite being observed to common variants, might in fact result from causal mutations that are rare. One prediction of this hypothesis is that the allelic associations should be population-specific, as the causal mutations arose after the migrations that established different populations around the world. We selected 19 common variants found to be reproducibly associated to T2D risk in European populations and studied them in a large multiethnic case-control study (6,142 cases and 7,403 controls) among men and women from 5 racial/ethnic groups (European Americans, African Americans, Latinos, Japanese Americans, and Native Hawaiians). In analysis pooled across ethnic groups, the allelic associations were in the same direction as the original report for all 19 variants, and 14 of the 19 were significantly associated with risk. In summing the number of risk alleles for each individual, the per-allele associations were highly statistically significant (P

Translating a set of disease regions into insight about pathogenic mechanisms requires not only the ability to identify the key disease genes within them, but also the biological relationships among those key genes. Here we describe a statistical method, Gene Relationships Among Implicated Loci (GRAIL), that takes a list of disease regions and automatically assesses the degree of relatedness of implicated genes using 250,000 PubMed abstracts. We first evaluated GRAIL by assessing its ability to identify subsets of highly related genes in common pathways from validated lipid and height SNP associations from recent genome-wide studies. We then tested GRAIL, by assessing its ability to separate true disease regions from many false positive disease regions in two separate practical applications in human genetics. First, we took 74 nominally associated Crohn's disease SNPs and applied GRAIL to identify a subset of 13 SNPs with highly related genes. Of these, ten convincingly validated in follow-up genotyping; genotyping results for the remaining three were inconclusive. Next, we applied GRAIL to 165 rare deletion events seen in schizophrenia cases (less than one-third of which are contributing to disease risk). We demonstrate that GRAIL is able to identify a subset of 16 deletions containing highly related genes; many of these genes are expressed in the central nervous system and play a role in neuronal synapses. GRAIL offers a statistically robust approach to identifying functionally related genes from across multiple disease regions--that likely represent key disease pathways. An online version of this method is available for public use (http://www.broad.mit.edu/mpg/grail/).

Lymphoblastoid cell lines (LCLs), originally collected as renewable sources of DNA, are now being used as a model system to study genotype-phenotype relationships in human cells, including searches for QTLs influencing levels of individual mRNAs and responses to drugs and radiation. In the course of attempting to map genes for drug response using 269 LCLs from the International HapMap Project, we evaluated the extent to which biological noise and non-genetic confounders contribute to trait variability in LCLs. While drug responses could be technically well measured on a given day, we observed significant day-to-day variability and substantial correlation to non-genetic confounders, such as baseline growth rates and metabolic state in culture. After correcting for these confounders, we were unable to detect any QTLs with genome-wide significance for drug response. A much higher proportion of variance in mRNA levels may be attributed to non-genetic factors (intra-individual variance--i.e., biological noise, levels of the EBV virus used to transform the cells, ATP levels) than to detectable eQTLs. Finally, in an attempt to improve power, we focused analysis on those genes that had both detectable eQTLs and correlation to drug response; we were unable to detect evidence that eQTL SNPs are convincingly associated with drug response in the model. While LCLs are a promising model for pharmacogenetic experiments, biological noise and in vitro artifacts may reduce power and have the potential to create spurious association due to confounding.

Steroid hormones are believed to play an important role in prostate carcinogenesis, but epidemiological evidence linking prostate cancer and steroid hormone genes has been inconclusive, in part due to small sample sizes or incomplete characterization of genetic variation at the locus of interest. Here we report on the results of a comprehensive study of the association between HSD17B1 and prostate cancer by the Breast and Prostate Cancer Cohort Consortium, a large collaborative study. HSD17B1 encodes 17beta-hydroxysteroid dehydrogenase 1, an enzyme that converts dihydroepiandrosterone to the testosterone precursor Delta5-androsterone-3beta,17beta-diol and converts estrone to estradiol. The Breast and Prostate Cancer Cohort Consortium researchers systematically characterized variation in HSD17B1 by targeted resequencing and dense genotyping; selected haplotype-tagging single nucleotide polymorphisms (htSNPs) that efficiently predict common variants in U.S. and European whites, Latinos, Japanese Americans, and Native Hawaiians; and genotyped these htSNPs in 8,290 prostate cancer cases and 9,367 study-, age-, and ethnicity-matched controls. We found no evidence that HSD17B1 htSNPs (including the nonsynonymous coding SNP S312G) or htSNP haplotypes were associated with risk of prostate cancer or tumor stage in the pooled multiethnic sample or in U.S. and European whites. Analyses stratified by age, body mass index, and family history of disease found no subgroup-specific associations between these HSD17B1 htSNPs and prostate cancer. We found significant evidence of heterogeneity in associations between HSD17B1 haplotypes and prostate cancer across ethnicity: one haplotype had a significant (p < 0.002) inverse association with risk of prostate cancer in Latinos and Japanese Americans but showed no evidence of association in African Americans, Native Hawaiians, or whites. However, the smaller numbers of Latinos and Japanese Americans in this study makes these subgroup analyses less reliable. These results suggest that the germline variants in HSD17B1 characterized by these htSNPs do not substantially influence the risk of prostate cancer in U.S. and European whites.