Several methods have been proposed to impute genotypes at untyped markers using observed genotypes and genetic data from a reference panel. We used the Genetic Analysis Workshop 16 rheumatoid arthritis case-control dataset to compare the performance of four of these imputation methods: IMPUTE, MACH, PLINK, and fastPHASE. We compared the methods' imputation error rates and performance of association tests using the imputed data, in the context of imputing completely untyped markers as well as imputing missing genotypes to combine two datasets genotyped at different sets of markers. As expected, all methods performed better for single-nucleotide polymorphisms (SNPs) in high linkage disequilibrium with genotyped SNPs. However, MACH and IMPUTE generated lower imputation error rates than fastPHASE and PLINK. Association tests based on allele "dosage" from MACH and tests based on the posterior probabilities from IMPUTE provided results closest to those based on complete data. However, in both situations, none of the imputation-based tests provide the same level of evidence of association as the complete data at SNPs strongly associated with disease.

Genetic Analysis Workshop 18 (GAW18) focused on identification of genes and functional variants that influence complex phenotypes in human sequence data. Data for the workshop were donated by the T2D-GENES Consortium and included whole genome sequences for odd-numbered autosomes in 464 key individuals selected from 20 Mexican American families, a dense set of single-nucleotide polymorphisms in 959 individuals in these families, and longitudinal data on systolic and diastolic blood pressure measured at 1-4 examinations over a period of 20 years. Simulated phenotypes were generated based on the real sequence data and pedigree structures. In the design of the simulation model, gene expression measures from the San Antonio Family Heart Study (not distributed as part of the GAW18 data) were used to identify genes whose mRNA levels were correlated with blood pressure. Observed variants within these genes were designated as functional in the GAW18 simulation if they were nonsynonymous and predicted to have deleterious effects on protein function or if they were noncoding and associated with mRNA levels. Two simulated longitudinal phenotypes were modeled to have the same trait distributions as the real systolic and diastolic blood pressure data, with effects of age, sex, and medication use, including a genotype-medication interaction. For each phenotype, more than 1000 sequence variants in more than 200 genes present on the odd-numbered autosomes individually explained less than 0.01-2.78% of phenotypic variance. Cumulatively, variants in the most influential gene explained 7.79% of trait variance. An additional simulated phenotype, Q1, was designed to be correlated among family members but to not be associated with any sequence variants. Two hundred replicates of the phenotypes were simulated, with each including data for 849 individuals.

The Genetic Analysis Workshops (GAW) are a forum for development, testing, and comparison of statistical genetic methods and software. Each contribution to the workshop includes an application to a specified data set. Here we describe the data distributed for GAW19, which focused on analysis of human genomic and transcriptomic data.

Genetic Analysis Workshop 17 (GAW17) provided a platform for evaluating existing statistical genetic methods and for developing novel methods to analyze rare variants that modulate complex traits. In this article, we present an overview of the 1000 Genomes Project exome data and simulated phenotype data that were distributed to GAW17 participants for analyses, the different issues addressed by the participants, and the process of preparation of manuscripts resulting from the discussions during the workshop.

The nuclear factor-kappaB (NF-kappaB) family of transcription factors regulates the expression of a variety of genes involved in apoptosis and immune response. We examined relationships between genotypes at five NF-kappaB subunits (NFKB1, NFKB2, REL, RELA, and RELB) and variable expression levels of 15 NF-kappaB regulated proteins with heritability greater than 0.40: BCL2A1, BIRC2, CD40, CD44, CD80, CFLAR, CR2, FAS, ICAM1, IL15, IRF1, JUNB, MYC, SLC2A5, and VCAM1. SNP genotypes and expression phenotypes from pedigrees of Utah residents with ancestry from northern and western Europe were provided by Genetic Analysis Workshop 15 and supplemented with additional genotype data from the International HapMap Consortium. We conducted association, linkage, and family-based association analyses between each candidate gene and the 15 heritable expression phenotypes. We observed consistent results in association and linkage analyses of the NFKB1 region (encoding p50) and levels of FAS and IRF1 expression. FAS is a cell surface protein that also belongs to the TNF-receptor family; signals through FAS are able to induce apoptosis. IRF1 is a member of the interferon regulatory transcription factor family, which has been shown to regulate apoptosis and tumor-suppression. Analyses in the REL region (encoding c-Rel) revealed linkage and association with CD40 phenotype. CD40 proteins belong to the tumor necrosis factor (TNF)-receptor family, which mediates a broad variety of immune and inflammatory responses. We conclude that variation in the genes encoding p50 and c-Rel may play a role in NF-kappaB-related transcription of FAS, IRF1, and CD40.

Several methods to identify tagging single-nucleotide polymorphisms (SNPs) are in common use for genetic epidemiologic studies; however, there may be loss of information when using only a subset of SNPs. We sought to compare the ability of commonly used pairwise, multimarker, and haplotype-based tagging SNP selection methods to detect known associations with quantitative expression phenotypes. Using data from HapMap release 21 on unrelated Utah residents with ancestors from northern and western Europe (CEPH-Utah, CEU), we selected tagging SNPs in five chromosomal regions using ldSelect, Tagger, and TagSNPs. We found that SNP subsets did not substantially overlap, and that the use of trio data did not greatly impact SNP selection. We then tested associations between HapMap genotypes and expression phenotypes on 28 CEU individuals as part of Genetic Analysis Workshop 15. Relative to the use of all SNPs (n = 210 SNPs across all regions), most subset methods were able to detect single-SNP and haplotype associations. Generally, pairwise selection approaches worked extremely well, relative to use of all SNPs, with marked reductions in the number of SNPs required. Haplotype-based approaches, which had identified smaller SNP subsets, missed associations in some regions. We conclude that the optimal tagging SNP method depends on the true model of the genetic association (i.e., whether a SNP or haplotype is responsible); unfortunately, this is often unknown at the time of SNP selection. Additional evaluations using empirical and simulated data are needed.

The synthetic association hypothesis proposes that common genetic variants detectable in genome-wide association studies may reflect the net phenotypic effect of multiple rare polymorphisms distributed broadly within the focal gene rather than, as often assumed, the effect of common functional variants in high linkage disequilibrium with the focal marker. In a recent study, Dickson and colleagues demonstrated synthetic association in simulations and in two well-characterized, highly polymorphic human disease genes. The converse of this hypothesis is that rare variant genotypes must be correlated with common variant genotypes often enough to make the phenomenon of synthetic association possible. Here we used the exome genotype data provided for Genetic Analysis Workshop 17 to ask how often, how well, and under what conditions rare variant genotypes predict the genotypes of common variants within the same gene. We found nominal evidence of correlation between rare and common variants in 21-30% of cases examined for unrelated individuals; this rate increased to 38-44% for related individuals, underscoring the segregation that underlies synthetic association.

No abstract available

Due to the growing need to combine data across multiple studies and to impute untyped markers based on a reference sample, several analytical tools for imputation and analysis of missing genotypes have been developed. Current imputation methods rely on single imputation, which ignores the variation in estimation due to imputation. An alternative to single imputation is multiple imputation. In this paper, we assess the variation in imputation by completing both single and multiple imputations of genotypic data using MACH, a commonly used hidden Markov model imputation method. Using data from the North American Rheumatoid Arthritis Consortium genome-wide study, the use of single and multiple imputation was assessed in four regions of chromosome 1 with varying levels of linkage disequilibrium and association signals. Two scenarios for missing genotypic data were assessed: imputation of untyped markers and combination of genotypic data from two studies. This limited study involving four regions indicates that, contrary to expectations, multiple imputations may not be necessary.

Microarray technologies allow the measurement of the expression levels of thousands of transcripts at the same time. As part of Genetic Analysis Workshop 15 (GAW15), we analyzed a data set that measured the expression of more than 3000 genes in 14 families. Our goal was to identify genomic regions that regulate the expression of several genes at the same time. We tried two different approaches: one was maximum likelihood-based variance-component linkage analysis and the other was a new linkage regression approach. We detected some loci that were linked with the expression level of more genes than would be expected by chance. These loci are candidates for master regulators of transcription (MRT). Finally, for each candidate MRT, we did a gene ontology (GO) analysis to test whether the genes linked to it were biologically related.

GAW20 provided a platform for developing and evaluating statistical methods to analyze human lipid-related phenotypes, DNA methylation, and single-nucleotide markers in a study involving a pharmaceutical intervention. In this article, we present an overview of the data sets and the contributions analyzing these data. The data, donated by the Genetics of Lipid Lowering Drugs and Diet Network (GOLDN) investigators, included data from 188 families (N = 1105) which included genome-wide DNA methylation data before and after a 3-week treatment with fenofibrate, single-nucleotide polymorphisms, metabolic syndrome components before and after treatment, and a variety of covariates. The contributions from individual research groups were extensively discussed prior, during, and after the Workshop in groups based on discussion themes, before being submitted for publication.

Joint analyses of correlated phenotypes in genetic epidemiology studies are common. However, these analyses primarily focus on genetic correlation between traits and do not take into account environmental correlation. We describe a method that optimizes the genetic signal by accounting for stochastic environmental noise through joint analysis of a discrete trait and a correlated quantitative marker. We conducted bivariate analyses where heritability and the environmental correlation between the discrete and quantitative traits were calculated using Genetic Analysis Workshop 17 (GAW17) family data. The resulting inverse value of the environmental correlation between these traits was then used to determine a new β coefficient for each quantitative trait and was constrained in a univariate model. We conducted genetic association tests on 7,087 nonsynonymous SNPs in three GAW17 family replicates for Affected status with the β coefficient fixed for three quantitative phenotypes and compared these to an association model where the β coefficient was allowed to vary. Bivariate environmental correlations were 0.64 (± 0.09) for Q1, 0.798 (± 0.076) for Q2, and -0.169 (± 0.18) for Q4. Heritability of Affected status improved in each univariate model where a constrained β coefficient was used to account for stochastic environmental effects. No genome-wide significant associations were identified for either method but we demonstrated that constraining β for covariates slightly improved the genetic signal for Affected status. This environmental regression approach allows for increased heritability when the β coefficient for a highly correlated quantitative covariate is constrained and increases the genetic signal for the discrete trait.

No abstract available

We compared results from linkage analyses of different phenotype measurements from the same gene expression traits and found that the strongest signals were detected by all expression measures that we considered. On average, that meant that the same quantitative trait loci (QTLs) were detected across methods, but the magnitude of the LOD score of each particular QTL and the false-positive ratio of QTL detection varied between the methods.

The application of pathway and gene-set based analyses to high-throughput data is increasingly common and represents an effort to understand underlying biology where single-gene or single-marker analyses have failed. Many such analyses rely on the a priori identification of genes associated with the trait of interest. In contrast, this variance-component-based approach creates a similarity matrix of individuals based on the expression of genes in each pathway.

The data set simulated for Genetic Analysis Workshop 17 was designed to mimic a subset of data that might be produced in a full exome screen for a complex disorder and related risk factors in order to permit workshop participants to investigate issues of study design and statistical genetic analysis. Real sequence data from the 1000 Genomes Project formed the basis for simulating a common disease trait with a prevalence of 30% and three related quantitative risk factors in a sample of 697 unrelated individuals and a second sample of 697 individuals in large, extended pedigrees. Called genotypes for 24,487 autosomal markers assigned to 3,205 genes and simulated affection status, quantitative traits, age, sex, pedigree relationships, and cigarette smoking were provided to workshop participants. The simulating model included both common and rare variants with minor allele frequencies ranging from 0.07% to 25.8% and a wide range of effect sizes for these variants. Genotype-smoking interaction effects were included for variants in one gene. Functional variants were concentrated in genes selected from specific biological pathways and were selected on the basis of the predicted deleteriousness of the coding change. For each sample, unrelated individuals and family, 200 replicates of the phenotypes were simulated.

Genetic Analysis Workshop 19 provided a platform for developing and evaluating statistical methods to analyze whole-genome sequence and gene expression data from a pedigree-based sample, as well as whole-exome sequence data from a large cohort of unrelated individuals. In this article we present an overview of the data sets, the GAW experience, and summaries of the contributions arranged into nine methodological themes.

The concept of breeding values, an individual's phenotypic deviation from the population mean as a result of the sum of the average effects of the genes they carry, is of great importance in livestock, aquaculture, and cash crop industries where emphasis is placed on an individual's potential to pass desirable phenotypes on to the next generation. As breeding or genetic values (as referred to here) cannot be measured directly, estimated genetic values (EGVs) are based on an individual's own phenotype, phenotype information from relatives, and, increasingly, genetic data. Because EGVs represent additive genetic variation, calculating EGVs in an extended human pedigree is expected to provide a more refined phenotype for genetic analyses. To test the utility of EGVs in genome-wide association, EGVs were calculated for 847 members of 20 extended Mexican American families based on 100 replicates of simulated systolic blood pressure. Calculations were performed in GAUSS to solve a variation on the standard Best Linear Unbiased Predictor (BLUP) mixed model equation with age, sex, and the first 3 principal components of sample-wide genetic variability as fixed effects and the EGV as a random effect distributed around the relationship matrix. Three methods of calculating kinship were considered: expected kinship from pedigree relationships, empirical kinship from common variants, and empirical kinship from both rare and common variants. Genome-wide association analysis was conducted on simulated phenotypes and EGVs using the additive measured genotype approach in the SOLAR software package. The EGV-based approach showed only minimal improvement in power to detect causative loci.

Many phenotypes may be influenced by the prenatal environment of the mother and/or maternal care, and these maternal effects may have a heritable component. We have implemented in the computer program SOLAR a variance components-based method for detecting indirect effects of maternal genotype on offspring phenotype. Of six phenotypes measured in three generations of the Framingham Heart Study, height showed the strongest evidence (P = 0.02) of maternal effect. We conducted a genome-wide association analysis for height, testing both the direct effect of the focal individual's genotype and the indirect effect of the maternal genotype. Offspring height showed suggestive evidence of association with maternal genotype for two single-nucleotide polymorphisms in the trafficking protein particle complex 9 gene TRAPPC9 (NIBP), which plays a role in neuronal NF-kappaB signalling. This work establishes a methodological framework for identifying genetic variants that may influence the contribution of the maternal environment to offspring phenotypes.

Gene identification using linkage, association, or genome-wide expression is often underpowered. We propose that formal combination of information from multiple gene-identification approaches may lead to the identification of novel loci that are missed when only one form of information is available.