Bayesian meta-analysis of genetic association studies with different sets of markers

Abstract

Robust assessment of genetic effects on quantitative traits or complex-disease risk requires synthesis of evidence from multiple studies. Frequently, studies have genotyped partially overlapping sets of SNPs within a gene or region of interest, hampering attempts to combine all the available data. By using the example of C-reactive protein (CRP) as a quantitative trait, we show how linkage disequilibrium in and around its gene facilitates use of Bayesian hierarchical models to integrate informative data from all available genetic association studies of this trait, irrespective of the SNP typed. A variable selection scheme, followed by contextualization of SNPs exhibiting independent associations within the haplotype structure of the gene, enhanced our ability to infer likely causal variants in this region with population-scale data. This strategy, based on data from a literature based systematic review and substantial new genotyping, facilitated the most comprehensive evaluation to date of the role of variants governing CRP levels, providing important information on the minimal subset of SNPs necessary for comprehensive evaluation of the likely causal relevance of elevated CRP levels for coronary-heart-disease risk by Mendelian randomization. The same method could be applied to evidence synthesis of other quantitative traits, whenever the typed SNPs vary among studies, and to assist fine mapping of causal variants.

Figure D1

Mean Pairwise LD Measures between Markers Used in the Simulation Study when Allowing LD Patterns to Vary across Studies

Figure 1

Location of the Eight CRP SNPs Typed Directly in the 26 Data Sets Included in This Study The upper track shows chromosomal location; the middle track shows SNP location and Log(P) for the per-allele random-effect meta-analysis (from left to right, the SNPs are ordered as follows: rs3093077, rs1205, rs1130864, rs1800947, rs1417938, rs3091244, rs2794521, and rs3093059); and the lower track shows the intron/exon structure of the CRP gene.

Figure 2

Graphical Representation of Equation (4) Solid and dotted lines represent stochastic and deterministic dependencies, respectively.

Figure 3

Pairwise LD Measures between Markers Used in the Simulation Study Pairwise LD Measures are r values.

Figure 4

Summary Effect from Traditional Meta-Analysis and Bayesian Multiple-SNP Hierarchical Linear Model of the Eight SNPs in the CRP Gene Values shown are additive genetic effects on (log) CRP levels with 95% confidence intervals or credible intervals for traditional and Bayesian analyses, respectively. For the Bayesian analysis, results are shown only for those markers that appear to be strongly associated after variable selection (see Figure 5). N/A refers to SNPs excluded from the model. The asterisk indicates the dominant model. Negative values indicate the variant allele is associated with a lower CRP concentration.

Figure 5

Results from the Multiple-SNP Meta-Analysis using the Bayesian Hierarchical Linear Model The shaded bars show the posterior probability that each SNP is included in a model, calculated from the posterior sample of models. The x axis indicates the additive effects of each SNP on log CRP plasma levels, conditional on that SNP being included in the model, and the y axis indicates the corresponding posterior density. The curves can thus be interpreted as smoothed histograms representing the probability that the SNP effects take the values on the x axis. Also shown are the densities, medians (▴), and 95% credible intervals (- - -) for the additive effects of each SNP on log CRP levels.

Figure 6

A Reduced Median Network Constructed with HapMap CEPH Data for a 20 kb Region Containing the CRP Gene Yellow circles indicate haplotypes. The size of each circle is proportional to the frequency of that haplotype in the HapMap CEPH population. Non-HapMap SNPs (indicated in italics) were placed on the network with information from other CEPH populations.

Figure 7

Genomic Context for CRP Gene (A) Ideogram depicting the chromosome and region in which the CRP gene lies (red line). (B) Gene diagram with introns and exons depicted as horizontal and vertical blue lines, respectively. (C) Pairwise r² LD values between independently associating SNPs from Bayesian analysis (identified in top left of window, position indicated by red arrow) and all other HapMap SNPs in the region (release 20, build 35, red = r² > 0.8, yellow = 0.5 < r² < 0.8, gray = 0.3 < r² < 0.5, blue = 0.2 < r² < 0.3, and dark gray = missing data).

Figure 8

Posterior Sample of Residuals from the Hierarchical Model of Material and Methods Fitted without the Between-Study Random-Effect Term μ_s