Computation of deregressed proofs for genomic selection when own phenotypes exist with an application in French show-jumping horses.

Abstract: Genomic evaluations often use as pseudo-phenotypes corrected means of progeny performances, like daughter yield deviations (DYD) in dairy species. In horse breeding, own performances are also available and performances from other relatives (as half sibs) may play an important part in the EBV because the number of progeny remains low, even for stallions. The first step for genomic selection in horses is therefore to generate pseudo-phenotypes for genomic analysis when parental or own information is considered. This work presents an easy method to compute deregressed EBV from regular pedigree-based genetic evaluations (EBV, reliabilities) to be used in genomic evaluations. The proposed methodology builds deregressed proofs so that they combine own performances (from genotyped individuals) and performances of relatives (outside of the genotyped sample). An application to show jumping horse data is presented. A sample of 908 stallions specialized in show jumping [71% Selle Français (SF), 17% foreign sport horses (FH), 13% Anglo Arab (AA)] were genotyped. Genotyping was performed using the Illumina Equine SNP50 BeadChip, and after quality tests, 44,444 SNP were retained. Two methods were used for genomic evaluation: GBLUP and BayesCπ, and 6 validation data sets were compared, chosen according to breeds SF + FH + AA or SF + FH, family structure (more than 3 half sibs), reliability of sires (>0.97) or sons (>0.72). In spite of a favorable genetic structure [linkage disequilibrium equal to 0.24 at 50 kb pairs], results showed low advantage of genomic evaluation. On the validation sample SF + FH + AA, the correlation between deregressed proofs and GBLUP or BayesCπ predictions was 0.39, 0.37, 0.51 according to the different validation data sets, compared with 0.36, 0.33, 0.53 obtained with BLUP predictions. Correlations were much lower on the SF + FH sample. Research is pursued to understand this low advantage of genomic selection and to improve the methodology for genomic evaluation in this context, which is less favorable than dairy cattle breeding.