Integration of estimates of SNP effects into a single-step genomic evaluation

Abstract

The aim of this research was to develop and validate a method that integrates estimates of single nucleotide polymorphism (SNP) effects and the associated prediction error (co)variance (PECs) matrix from a genomic evaluation into a single-step SNP Best Linear Unbiased Prediction (ssSNPBLUP) evaluation. As the PEC matrix is a dense matrix, the developed method was also tested with two different chromosome-wise matrices (that is, ignoring off-diagonal elements among chromosomes), and with a prediction error variance matrix (that is, ignoring all off-diagonal elements of the PEC matrix). Using simulated data from two dairy cattle populations with a genetic correlation between their traits of 0.80, we compared the genomic enhanced breeding values (GEBVs) predicted by the different integration methods to those of a joint ssSNPBLUP evaluation of both populations. The developed method, using the whole PEC matrix, resulted in GEBVs for selection candidates highly correlated and consistent with those from the joint ssSNPBLUP evaluation. Ignoring off-diagonal elements among chromosomes resulted in similar accurate results, but ignoring all PECs resulted in biased GEBVs in comparison to those of the joint evaluation. Therefore, an accurate integration of estimates of SNP effects and the associated PEC matrix into a single-step genomic evaluation is feasible and accurate when PEC of SNP effects within chromosomes are at least considered. The developed method can be readily implemented in existing software that support ssSNPBLUP models and can be adapted for single-step genomic BLUP models, though further research is needed to address potential computational challenges with these models.