Increasing the power of meta-analysis of genome-wide association studies to detect heterogeneous effects

Abstract

MotivationMeta-analysis is essential to combine the results of genome-wide association studies (GWASs). Recent large-scale meta-analyses have combined studies of different ethnicities, environments and even studies of different related phenotypes. These differences between studies can manifest as effect size heterogeneity. We previously developed a modified random effects model (RE2) that can achieve higher power to detect heterogeneous effects than the commonly used fixed effects model (FE). However, RE2 cannot perform meta-analysis of correlated statistics, which are found in recent research designs, and the identified variants often overlap with those found by FE.ResultsHere, we propose RE2C, which increases the power of RE2 in two ways. First, we generalized the likelihood model to account for correlations of statistics to achieve optimal power, using an optimization technique based on spectral decomposition for efficient parameter estimation. Second, we designed a novel statistic to focus on the heterogeneous effects that FE cannot detect, thereby, increasing the power to identify new associations. We developed an efficient and accurate p-value approximation procedure using analytical decomposition of the statistic. In simulations, RE2C achieved a dramatic increase in power compared with the decoupling approach (71% vs. 21%) when the statistics were correlated. Even when the statistics are uncorrelated, RE2C achieves a modest increase in power. Applications to real genetic data supported the utility of RE2C. RE2C is highly efficient and can meta-analyze one hundred GWASs in one day.Availability and implementationThe software is freely available at http://software.buhmhan.com/RE2C.Supplementary information Supplementary data are available at Bioinformatics online.

GWAS of five gynecologic diseases and cross-trait analysis in Japanese

“…Joint analysis of all cases versus shared controls was executed under the logistic regression model using mach2dat, and was compared with random-effect metaanalysis of mach2dat results of each disease using RE2C (v1.04) [27] to correct for the overlapping samples and increase the power for detection. The same set of analyses was also conducted under the linear mixed model using BOLT-LMM and compared with the corresponding mach2dat results.…”

Combining genome-wide studies of breast, prostate, ovarian and endometrial cancers maps cross-cancer susceptibility loci and identifies new genetic associations

“…Antagonistic effects across cancer types at the same allele also has important implications for the use of such alleles in the development and application of polygenic risk scores that have so far been confined to a consideration of allelic associations with a single cancer. The identification of alleles with opposite effects across the cancers was helped by that fact that the Han and Eskin model 11,12 used in this study can detect effects under heterogeneity in contrast to the standard fixed-effects model used in the 2016 meta-analysis 6 . Opposite associations across related diseases at the same allele are also seen in autoimmune 63 and psychiatric 64 disease groups.…”

“…We used meta-analysis based on the Han and Eskin model 11,12 to combine summary results for 9,530,997 variants with minor allele frequency > 1% from the largest genome-wide association data sets for susceptibility to breast 7 , prostate 8 , ovarian 9 , and endometrial cancers 10 published as of August 2019 (Methods). All cases and controls included were of European ancestry ( Supplementary Table 1).…”

On set-based association tests: insights from a regression using summary statistics

“…In that context, a likelihood ratio test was implemented for a mixed-effect model, and the resulting test is also known as the new RE meta-analysis. The original test of Han & Eskin (2011) was designed for meta-analyses of independent studies, and a modified procedure has since been developed by Lee, Eskin & Han (2017) to account for correlations between studies but without adjusting for covariate effects. Comparisons between the two approaches for meta-analyses and other studies warrant future investigations.…”