Dataset Information

A simple Bayesian mixture model with a hybrid procedure for genome-wide association studies.

ABSTRACT: Genome-wide association studies often face the undesirable result of either failing to detect any influential markers at all because of a stringent level for testing error corrections or encountering difficulty in quantifying the importance of markers by their P-values. Advocates of estimation procedures prefer to estimate the proportion of association rather than test significance to avoid overinterpretation. Here, we adopt a Bayesian hierarchical mixture model to estimate directly the proportion of influential markers, and then proceed to a selection procedure based on the Bayes factor (BF). This mixture model is able to accommodate different sources of dependence in the data through only a few parameters. Specifically, we focus on a standardized risk measure of unit variance so that fewer parameters are involved in inference. The expected value of this measure follows a mixture distribution with a mixing probability of association, and it is robust to minor allele frequencies. Furthermore, to select promising markers, we use the magnitude of the BF to represent the strength of evidence in support of the association between markers and disease. We demonstrate this procedure both with simulations and with SNP data from studies on rheumatoid arthritis, coronary artery disease, and Crohn's disease obtained from the Wellcome Trust Case-Control Consortium. This Bayesian procedure outperforms other existing methods in terms of accuracy, power, and computational efficiency. The R code that implements this method is available at http://homepage.ntu.edu.tw/~ckhsiao/Bmix/Bmix.htm.

SUBMITTER: Wei YC

PROVIDER: S-EPMC2987394 | biostudies-literature | 2010 Aug

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

A simple Bayesian mixture model with a hybrid procedure for genome-wide association studies.

Wei Yu-Chung YC Wen Shu-Hui SH Chen Pei-Chun PC Wang Chih-Hao CH Hsiao Chuhsing K CK

European journal of human genetics : EJHG 20100421 8

Genome-wide association studies often face the undesirable result of either failing to detect any influential markers at all because of a stringent level for testing error corrections or encountering difficulty in quantifying the importance of markers by their P-values. Advocates of estimation procedures prefer to estimate the proportion of association rather than test significance to avoid overinterpretation. Here, we adopt a Bayesian hierarchical mixture model to estimate directly the proporti ...[more]

PMID: 20407469

Similar Datasets

Project description:Characterizing the distribution of effects from genome-wide genotyping data is crucial for understanding important aspects of the genetic architecture of complex traits, such as number or proportion of non-null loci, average proportion of phenotypic variance explained per non-null effect, power for discovery, and polygenic risk prediction. To this end, previous work has used effect-size models based on various distributions, including the normal and normal mixture distributions, among others. In this paper we propose a scale mixture of two normals model for effect size distributions of genome-wide association study (GWAS) test statistics. Test statistics corresponding to null associations are modeled as random draws from a normal distribution with zero mean; test statistics corresponding to non-null associations are also modeled as normal with zero mean, but with larger variance. The model is fit via minimizing discrepancies between the parametric mixture model and resampling-based nonparametric estimates of replication effect sizes and variances. We describe in detail the implications of this model for estimation of the non-null proportion, the probability of replication in de novo samples, the local false discovery rate, and power for discovery of a specified proportion of phenotypic variance explained from additive effects of loci surpassing a given significance threshold. We also examine the crucial issue of the impact of linkage disequilibrium (LD) on effect sizes and parameter estimates, both analytically and in simulations. We apply this approach to meta-analysis test statistics from two large GWAS, one for Crohn's disease (CD) and the other for schizophrenia (SZ). A scale mixture of two normals distribution provides an excellent fit to the SZ nonparametric replication effect size estimates. While capturing the general behavior of the data, this mixture model underestimates the tails of the CD effect size distribution. We discuss the implications of pervasive small but replicating effects in CD and SZ on genomic control and power. Finally, we conclude that, despite having very similar estimates of variance explained by genotyped SNPs, CD and SZ have a broadly dissimilar genetic architecture, due to differing mean effect size and proportion of non-null loci.

Project description:BackgroundRecently mixed linear models are used to address the issue of "missing" heritability in traditional Genome-wide association studies (GWAS). The models assume that all single-nucleotide polymorphisms (SNPs) are associated with the phenotypes of interest. However, it is more common that only a small proportion of SNPs have significant effects on the phenotypes, while most SNPs have no or very small effects. To incorporate this feature, we propose an efficient Hierarchical Bayesian Model (HBM) that extends the existing mixed models to enforce automatic selection of significant SNPs. The HBM models the SNP effects using a mixture distribution of a point mass at zero and a normal distribution, where the point mass corresponds to those non-associative SNPs.ResultsWe estimate the HBM using Gibbs sampling. The estimation performance of our method is first demonstrated through two simulation studies. We make the simulation setups realistic by using parameters fitted on the Framingham Heart Study (FHS) data. The simulation studies show that our method can accurately estimate the proportion of SNPs associated with the simulated phenotype and identify these SNPs, as well as adapt to certain model mis-specification than the standard mixed models. In addition, we analyze data from the FHS and the Health and Retirement Study (HRS) to study the association between Body Mass Index (BMI) and SNPs on Chromosome 16, and replicate the identified genetic associations. The analysis of the FHS data identifies 0.3% SNPs on Chromosome 16 that affect BMI, including rs9939609 and rs9939973 on the FTO gene. These two SNPs are in strong linkage disequilibrium with rs1558902 (Rsq =0.901 for rs9939609 and Rsq =0.905 for rs9939973), which has been reported to be linked with obesity in previous GWAS. We then replicate the findings using the HRS data: the analysis finds 0.4% of SNPs associated with BMI on Chromosome 16. Furthermore, around 25% of the genes that are identified to be associated with BMI are common between the two studies.ConclusionsThe results demonstrate that the HBM and the associated estimation algorithm offer a powerful tool for identifying significant genetic associations with phenotypes of interest, among a large number of SNPs that are common in modern genetics studies.

Dataset Information

A simple Bayesian mixture model with a hybrid procedure for genome-wide association studies.

Publications

A simple Bayesian mixture model with a hybrid procedure for genome-wide association studies.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets