Use of Single Step GBLUP for Genomic Predictions and Genome-wide Association Studies

A single-step GBLUP (ssGBLUP) is a procedure that calculates GEBVs based on combined pedigree, genomic and phenotypic information. The procedure achieves these goals by blending traditional pedigree relationships with those derived from genetic markers. In practical applications involving national data sets in cattle, pigs, chicken and sheep, ssGBLUP exhibited superior accuracy with simplicity of application and inexpensive computing.  However, the applications pointed to general problems of genomic predictions where GEBV of particular group of animals may be biased upwards or downwards depending on properties of the genomic relationship matrix (GRM). Biases are minimized or removed when such a matrix accounts for admixed populations and the genetic distance of the genotyped population to the base population. ssGBLUP can be modified for genome-wide association analysis (GWAS). For GWAS, GEBVs are computed initially using a GRM derived assuming equal weight per SNP. Then GEBVs are converted to SNP effects, which are used to estimate SNP weights. Iteration improves identification of top SNP markers and increases the accuracy of GEBV. In practical comparisons, Manhattan plots of Classical GWAS and ssGBLUP were very similar while plots by BayesB were dominated by a single region. The largest GWAS with ssGBLUP involved a reaction norm model for climate-change research with data involving about 100 million test days for 10 million of Holsteins and 5000 genotypes. The analysis took about 12 hrs. ssGBLUP is implemented in BLUPF90 package available at nce.ads.uga.edu with modules for BLUP, variance component estimation using REML and Bayesian models, and GWAS.