Genome Wide Association Analyses Based on Broadly Different Specifications for Prior Distributions, Genomic Windows, and Estimation Methods

Abstract

17A popular strategy (EMMAX) for genome wide association (GWA) analysis fits all marker 18 effects as classical random effects (i.e., Gaussian prior) by which association for the specific 19 marker of interest is inferred by treating its effect as fixed. It seems more statistically coherent to 20 specify all markers as sharing the same prior distribution, whether it is Gaussian, heavy-tailed 21 (BayesA), or has variable selection specifications based on a mixture of, say, two Gaussian 22 SUMMARY 37 Genome wide association (GWA) analyses strategies have been improved by 38 simultaneously fitting all marker effects when inferring upon any single marker effect, with the 39 most popular distributional assumption being normality. Using data generated from 43,266 40 genotypes on 922 Duroc-Pietrain F2 cross pigs, we demonstrate that GWA studies could 41 particularly benefit from more flexible heavy-tailed or variable selection distributional 42 assumptions. Furthermore, these associations should not just be based on single markers or even 43 genomic windows of markers of fixed physical distances (0.5 -3.0 Mb) but based on adaptively 44 determined genomic windows using linkage disequilibrium information. 45

A certain invariance property of BLUE in a whole‐genome regression context

“…The results presented in this paper, shown first by Gianola et al. () just for GLS (BLUE), are seemingly unrecognized in the GWAS literature (e.g., Chen, Steibel, & Tempelman, ). An additional and probably useful result reported here is that represented by Equation : the variance of the estimate of any of the marker effects tested in GWAS can be obtained via a simple adjustment of the variance obtained with all markers entering into the similarity matrix.…”

Genome-wide association analyses based on a multiple-trait approach for modeling feed efficiency

“…A genomic window approach was considered given that it has been increasingly utilized to help alleviate the effects of highly multicollinear SNP genotypes (i.e., in high LD with each other) on GWA inferences. Recently, Chen et al (2017) investigated the use of adaptively determined genomic windows based on LD patterns to mitigate issues where a large LD block is shared between one or more fixedsized regions. They determined the utility of such an adaptive approach to be greater for Bayesian-based GWA methods but less so for EMMAX-like approaches such as those used in this paper.…”

“…such that g ir is of dimension n r × 1. A joint test for all n r markers within window r has been proposed by Chen et al (2017) and involves computing the test statistic in Equation [6]:…”

“…is equivalent to conducting GLS inference on g ir , treating all other SNP effects as random. That is, by extending BernalRubio et al (2016),Chen et al (2017) demonstrated that χ r 2 is distributed approximately as a chi-squared random variable with n r degrees of freedom under the null hypothesis H o : g ir = 0. Data were collected from 6,937 Holstein cows involving 16 research stations or major studies from Scotland, the Netherlands, Canada, and the United States.…”

A Multiple-trait Bayesian Variable Selection Regression Method for Integrating Phenotypic Causal Networks in Genome-Wide Association Studies

“…Given the MCMC samples of q w , the window posterior probability of association (WPPA) is calculated as the proportion of MCMC samples of q w that exceed a specific value T (Fernando and Garrick 2013; Chen et al 2017; Lloyd-Jones et al 2017). In this paper, associations are tested for non-overlapping windows of 100 SNPs, and genomic windows that explain over of the total genetic variance were deemed to be of potential interest (i.e., , where N is the total number of windows).…”