Improvement of genomic prediction accuracy for residual feed intake by prioritizing genetic markers identified by genome-wide association and gene expression

Abstract

<p>Genomic selection strategies applied to complex traits like residual feed intake (RFI) could improve the selection for feed efficiency in livestock. In the last decade, more information about the undelying genetic architecture for traits like RFI have been obtained through genomic wide association studies (GWAS) and transcriptomic studies. The aim of this study was to test whether combining information from GWAS and gene expression significantly associated (GSA) results could improve the accuracy of genomic prediction. We evaluated the gain in accuracy of prediction of RFI in 2190 Angus steers using medium-density and high-density SNP panels (770k) and by adding pre-selected SNPs (top SNPs) most significant in GWAS and close GSA from a gene expression. Two cross-validation designs were compared, one where the same dataset was used for training and GWAS discovery (4CV) and one where the discovery was separated from the training set (4x4CV). There was no improvement in prediction when using 770k compared with the medium density SNP panel. The 4x4CV design increase in accuracy by 1.2 and 2.7 percent point when top-SNPs (-log10(P)=3.5) were used, compared to using only 50k or 770k, respectively. The 4CV design showed lower accuracy when using top SNPs and the predictions were much more biased. The use of top SNPs in combination with selected SNPs located inside GSA reduced the bias in prediction compared with using only top SNPs in 4x4CV and slightly increased the accuracy of prediction. Genomic prediction accuracy can be improved when using selected SNPs from GWAS and GSA.</p>