Comparison of Two Meta-Analysis Methods: Inverse-Variance-Weighted Average and Weighted Sum of Z-Scores

Lee, Cue Hyunkyu; Cook, Seungho; Lee, Ji Sung; Han, Buhm

doi:10.5808/gi.2016.14.4.173

Cited by 208 publications

(142 citation statements)

References 26 publications

Supporting

Mentioning

142

Contrasting

Order By: Relevance

“…We extracted P values, estimated SNP effects and standard errors at each scanning locus. We considered two popular metaanalysis approaches: the inverse variance-weighted average and the weighted sum of z-scores [75][76][77] . For weighted sum of z-scores, we tried two weighing schemes, i.e., the sample size and squared root of the sample size, and found the results were very similar, and were slightly better than that of the inverse variance-weighted average.…”

Section: Meta-analysis In the Lgsm Ail Micementioning

confidence: 99%

Genome-wide Associations Reveal Human-Mouse Genetic Convergence and Modifiers of Myogenesis, CPNE1 and STC2

Cordero

Gonzales

Parker

et al. 2019

The American Journal of Human Genetics

View full text Add to dashboard Cite

Muscle bulk in adult healthy humans is highly variable even after accounting for height, age and sex. Low muscle mass, due to fewer and/or smaller constituent muscle fibers, would exacerbate the impact of muscle loss occurring in aging or disease. Genetic variability substantially influences muscle mass differences, but causative genes remain largely unknown. In a genome-wide association study (GWAS) on appendicular lean mass (ALM) in a population of 85,750 middle-age (38-49 years) individuals from the UK Biobank (UKB) we found 182 loci associated with ALM (P<5x10-8). We replicated associations for 78% of these loci (P<5x10-8) with ALM in a population of 181,862 elderly (60-74 years) individuals from UKB. We also conducted a GWAS on hindlimb skeletal muscle mass of 1,867 mice from an advanced intercross between two inbred strains (LG/J and SM/J) which identified 23 quantitative trait loci. 38 positional candidates distributed across 5 loci overlapped between the two species. In vitro studies of positional candidates confirmed CPNE1 and STC2 as modifiers of myogenesis. Collectively, these findings shed light on the genetics of muscle mass variability in humans and identify targets for the development of interventions for treatment of muscle loss. The overlapping results between humans and the mouse model GWAS point to shared genetic mechanisms across species.

show abstract

Section: Meta-analysis In the Lgsm Ail Micementioning

confidence: 99%

Genome-wide Associations Reveal Human-Mouse Genetic Convergence and Modifiers of Myogenesis, CPNE1 and STC2

Cordero

Gonzales

Parker

et al. 2019

The American Journal of Human Genetics

View full text Add to dashboard Cite

show abstract

“…The inverse-variance-weighted effect-size method ( Cochran, 1954 ; de Bakker et al , 2008 ; Fleiss, 1993 ; Mantel and Haenszel, 1959 ) and the weighted sum-of-z-scores method ( de Bakker et al , 2008 ; Han and Eskin, 2011 ; Zaykin, 2011 ) are used widely. We only describe the former, because the two methods are approximately equivalent ( Lee et al , 2016 ). Let

be the effect-size estimates, such as log odds ratios or regression coefficients, in

independent studies.…”

Section: Methodsmentioning

confidence: 99%

“…The inverse-variance-weighted effect-size estimator is the sum of

weighted with weights

The variance of

It follows that the standard error of

. Note that

is minimized only if the weights are inverse variances, which explains the method’s name ( Cochran, 1954 ; Greene, 2012 ; Lee et al , 2016 ). We can then build a summary z-score,

which follows

under the null hypothesis of no association

0

.…”

Section: Methodsmentioning

confidence: 99%

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

2017

View full text Add to dashboard Cite

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.

show abstract

“…So that the joint likelihood, as factorized in Eq 48, is simply . As shown previously [Lee et al, 2016], maximizing the approximate joint likelihood is equivalent to inverse-variance meta-analysis, which takes the form …”

Section: Qtl Mapping Proceduresmentioning

confidence: 99%

Scalable unified framework of total and allele-specific counts for cis-QTL, fine-mapping, and prediction

Liang

Aguet

Barbeira

et al. 2020

Preprint

View full text Add to dashboard Cite

9Genome-wide association studies (GWAS) have been highly successful in identifying genomic 10 loci associated with complex traits. However, identification of the causal genes that mediate 11 these associations remains challenging, and many approaches integrating transcriptomic data 12 with GWAS have been proposed. However, there currently exist no computationally scalable 13 methods that integrate total and allele-specific gene expression to maximize power to detect 14 genetic effects on gene expression. Here, we describe a unified framework that is scalable to 15 studies with thousands of samples. Using simulations and data from GTEx, we demonstrate 16 an average power gain equivalent to a 29% increase in sample size for genes with sufficient 17 allele-specific read coverage. We provide a suite of freely available tools, mixQTL, mixFine, and 18 mixPred, that apply this framework for mapping of quantitative trait loci, fine-mapping, and 19 prediction. 20 1 Introduction 21 Genome-wide association studies (GWAS) have identified tens of thousands of genomic loci asso-22 ciated with complex traits. A large majority of these loci lie in non-coding regions of the genome, 23 1 which hinders identification of the underlying molecular mechanisms and causal genes. Multi-24 ple methods have been developed to integrate GWAS results with expression quantitatite trait 25 loci (eQTLs), to test whether complex trait associations are mediated through regulation of gene 26 expression. Two strategies are commonly employed: 1) association-based approaches including 27 PrediXcan [Gamazon et al., 2015], fusion [Gusev et al., 2016], and smr [Zhu et al., 2016]; and 28 2) colocalization-based approaches including coloc [Giambartolomei et al., 2014], eCAVIAR [Hor-29 mozdiari et al., 2016], and enloc [Wen et al., 2017]. These approaches rely on high-quality eQTL 30 mapping, fine-mapping, and gene expression predictions. 31 In cis-eQTL analysis, allele-specific expression (ASE), i.e., the relative expression difference 32 between the two haplotypes, captures the genetic effect of nearby variants. ASE provides additional 33 signal to total read count, and several methods have been proposed to combine total and allele-34 specific read count for QTL mapping, such as TReCASE [Sun, 2012], WASP [Van De Geijn et al., 35 2015], and RASQUAL [Kumasaka et al., 2016]). However, these methods are computationally too 36 costly to be applied to sample sizes beyond a few hundred and as a result have not been applied to 37 large-scale studies like GTEx, which includes over 17,000 samples across 49 tissues. Recently, two 38 fine-mapping approaches have been proposed utilizing effect size estimates obtained from both ASE 39 and eQTL mapping via meta-analysis [Zou et al., 2019; Wang et al., 2020]. However, no existing 40 methods, to our knowledge, provides a unified framework of total and allele-specific counts with 41 explicit multi-SNP modeling for QTL mapping, fine-mapping, and prediction. 42 By assuming a log-linear model for transcript expression level...

show abstract

Comparison of Two Meta-Analysis Methods: Inverse-Variance-Weighted Average and Weighted Sum of Z-Scores

Cited by 208 publications

References 26 publications

Genome-wide Associations Reveal Human-Mouse Genetic Convergence and Modifiers of Myogenesis, CPNE1 and STC2

Genome-wide Associations Reveal Human-Mouse Genetic Convergence and Modifiers of Myogenesis, CPNE1 and STC2

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

Scalable unified framework of total and allele-specific counts for cis-QTL, fine-mapping, and prediction

Contact Info

Product

Resources

About