|Chen, Chunyu - Michigan State University|
|Weigel, Kent - University Of Wisconsin|
|Spurlock, Diane - Iowa State University|
|Vandehaar, Michael - Michigan State University|
|Staples, Charles - University Of Florida|
|Tempelman, Robert - Michigan State University|
Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 2/28/2017
Publication Date: 6/24/2017
Citation: Chen, C., Weigel, K.A., Connor, E.E., Spurlock, D.M., Vandehaar, M.J., Staples, C.R., Tempelman, R.J. 2017. Bayesian whole-genome prediction and genome-wide association analysis with missing genotypes using variable selection. Journal of Dairy Science. 100(Suppl. 2):412 (abstr. 469).
Technical Abstract: Single-step Genomic Best Linear Unbiased Predictor (ssGBLUP) has become increasingly popular for whole-genome prediction (WGP) modeling as it utilizes any available pedigree and phenotypes on both genotyped and non-genotyped individuals. The WGP accuracy of ssGBLUP has been demonstrated to be greater than or equivalent to popular Bayesian regression models. However, these assessments have not typically included phenotypes on non-genotyped individuals in the Bayesian regression analyses, making the interpretation of these comparisons difficult. Increasingly, ssGBLUP has been used for genome-wide association (GWA) studies, although there is no clear guidance on how to determine statistical significance in these analyses. We addressed this issue by proposing a GWA based on a Bayesian single-step stochastic search and variable selection (ssSSVS) model that allows for phenotypes on non-genotyped animals. Our study was based on a dairy consortium dataset including 3,186 Holstein cows from 6 US research stations using the 60671 USDA-ARS bovine SNP panel. In a simulation study using these same genotypes, different numbers of causal variants (nc = 30, 300, or 3,000) were randomly assigned to the markers, masking 20% of cows as non-genotyped, for a trait having a heritability of 0.25. We determined that ssSSVS had greater (P<0.05) WGP accuracy than ssGBLUP with nc = 30 or nc = 300. Moreover, ssSSVS always performed better (P<0.05) than ssGBLUP for GWA measured as partial area under a receiver-operating characteristic (ROC) curve (pAUC) up to a false positive rate of 5%. In a 10-fold within-station cross-validation study using phenotypes from the dairy consortium, we determined that ssSSVS had greater (P<0.05) WGP accuracies in milk fat compared to ssGBLUP for genotyped individuals, although no such differences were detected for body weight; we also found no significant difference between ssSSVS and ssGBLUP for WGP accuracies for non-genotyped individual for both traits. Overall, we find ssSSVS to be a promising method for both WGP and GWA, particularly for genetic architectures characterized by a few genes with large effects.