Rapid variance components–based method for whole-genome association analysis

Abstract

The variance component tests used in genome-wide association studies (GWAS) including large sample sizes become computationally exhaustive when the number of genetic markers is over a few hundred thousand. We present an extremely fast variance components-based two-step method, GRAMMAR-Gamma, developed as an analytical approximation within a framework of the score test approach. Using simulated and real human GWAS data sets, we show that this method provides unbiased estimates of the SNP effect and has a power close to that of the likelihood ratio test-based method. The computational complexity of our method is close to its theoretical minimum, that is, to the complexity of the analysis that ignores genetic structure. The running time of our method linearly depends on sample size, whereas this dependency is quadratic for other existing methods. Simulations suggest that GRAMMAR-Gamma may be used for association testing in whole-genome resequencing studies of large human cohorts.

10 Years of GWAS Discovery: Biology, Function, and Translation

“…New Analysis Methodology Underpinning New Discovery GWAS data have led to new analysis methods that fall into a number of categories depending on their purpose: (1) methods of better modeling population structure and relatedness between individuals in a sample during association analyses, [28][29][30][31][32][33][34] (2) methods of detecting novel variants and gene loci on the basis of GWAS summary statistics, [35][36][37] (3) methods of estimating and partitioning genetic (co)variance, 38,39 and (4) methods of inferring causality. [40][41][42] In addition, GWAS discoveries and interpretation have benefited substantially from improved algorithms in statistical imputation of unobserved genotypes and statistical imputation of human leukocyte antigen (HLA) genes and amino acid polymorphisms.…”

GWAPP: A Web Application for Genome-Wide Association Mapping in Arabidopsis

“…The only statistical method that appears to be effective for this purpose in A. thaliana is a mixed model that takes population structure into account using a genetic relatedness matrix (Yu et al, 2006;Zhao et al, 2007). Software that implements these models exists (Bradbury et al, 2007;Kang et al, 2010;Zhang et al, 2010;Lipka et al, 2011;Lippert et al, 2011;Zhou and Stephens, 2012;Svishcheva et al, 2012), but requires the user to provide both the genotype and phenotype data, as well as filtering and ordering the data appropriately. In addition, they provide little or no help in analyzing the results.…”

Genetic Studies: The Linear Mixed Models in Genome-wide Association Studies

“…The pressure on computation leads to the development of the score test based variance components methods (Fast association score test based analysis (FASTA [50]) and genome-wide rapid association mixed model and regression with GRAMMAR Gamma factor (GRAMMAR-Gamma [38])). FASTA is a two-stage approach where the population parameters and genetic similarity matrix are estimated once in the first step, and the score test with the previously computed population parameters and kinship matrix are applied to each marker to detect its effect, moreover, the likelihood ratio test is applied again to few candidate markers from previous score test to achieve more accurate significance in the second step.…”

“…However, variance components models cannot be applied to the genomic data with millions of SNPs and thousands of individuals owing to the heavy burden on estimating random parameters. Motivated by this limitation of variance components model, some efficient LMM-based approaches in GWAS have been proposed (Yu's model [23], compressed MLM with P3D [35], EMMAX [36], FaST-LMM [37] and GRAMMAR-Gamma [38]) recently. The model of linear mixed model based methods is set as,…”