Genotype by environment interaction for productive and reproductive traits in beef cattle using imputed whole genome sequence.

Abstract

Accounting for genotype by environment interaction (GxE) and using genomic information may enhance the prediction accuracy $\left(\widehat{\text{ACC}}\right)$ of breeding values. Hence, the objective of this study was to evaluate the gain in $\widehat{\text{ACC}}$ using single-step genomic BLUP using high-density SNP chip (ssGBLUP_HD) or whole genome imputed sequence (ssGBLUP_SEQ) compared to pedigree BLUP in the presence of GxE. Phenotypic data for age at first calving (AFC), scrotal circumference (SC), post-weaning weight gain (PWG), and yearling weight (YW) were obtained from commercial breeding programs of Nellore cattle. There were 1,578,591 animals in the pedigree, from which 51,485 had genotypes with high-density SNP chip (HD) and whol- genome imputed sequence (WGS), totaling 460,578 and 2,437,948 SNPs, respectively, after quality control. Contemporary group effects, estimated with a regular animal model (without modeling GxE), were used to define the environmental gradients (EG) for the reaction norm model (RNM). Genetic sensitivity to environmental variation was assessed by fitting three different linear RNM: the first considering only pedigree (BLUP), the second also considering the genomic information from HD, and the third considering the genomic information from WGS. The validation was carried out for genotyped young bulls, with no progeny records in the reduced data and at least one in the complete data. Models were compared using prediction accuracy, dispersion, correlation between the breeding values from reduced data and complete data, and bias from the linear regression method. Re-ranking between animals and heterogeneity of genetic variance in different EG were observed, suggesting the presence of GxE. The results for the regression coefficients of the RNM showed, in general, that the inclusion of genomic information increased the $\widehat{\text{ACC}}$ for the RNM regression coefficients for all traits. For SC, PWG, and YW, the highest accuracies were obtained with ssGBLUP_SEQ. Conversely, AFC had higher accuracy with ssGBLUP_HD. In addition, the $\widehat{\text{ACC}}$ for genotyped young bulls increased as the EG increased. In conclusion, ssGBLUP_SEQ yielded higher $\widehat{\text{ACC}}$ and correlation and a lower bias than the BLUP across all EG, indicating that the implementation of genomic selection using the whole genome sequence and accounting for GxE benefits this Nellore beef cattle population.