A Genomic Prediction Cross-Validation Approach Combining Ewe Repeated Phenotypes and Ram Daughter Trait Deviations

Abstract

Reproduction traits generally are expressed later in life and have high economic value, which makes them good candidates for improvement through genomic selection. We derive genomic prediction equations for number of lambs weaned (NLW) from genotyped rams' daughter records and repeated ewe records for genotyped ewes to make the best use of all available information. In order to assess the accuracy of the genomic estimated breeding values (GEBV) for this trait, an across sire family cross-validation design is proposed, and was compared with fully random cross-validation. The accuracy of genomic BLUP (GBLUP) using both ram and ewe records is compared with GBLUP using only ewe records, as well as BLUP with no genomic information using both ram and ewe records. The combined approach resulted in higher accuracies than both GBLUP only using ewe records, and BLUP. The approach provides a way to make use of all data available to maximise accuracy of GEBVs for traits such as NLW.