Genomic evaluation models including casein and/or SNP50K information in French dairy goat population

International audience

anglais. Genomic selection using single-step GBLUP (ssGBLUP) approach (based on female phenotypes) will be implemented at the end of 2017 in the two main French dairy goat breeds (Alpine and Saanen). Higher accuracy is expected by implementing approaches that include well-known major genes such as the αs1 casein gene. The αs1 casein is a multi-allelic gene located on chromosome 6 with significant effect on protein content in the two goat breeds. 6 alleles were identified in French goat populations and were each associated with a different level of αs1 casein synthesis. Genotypes of αs1 casein gene were available since 1982 for 4,203 bucks used for artificial insemination and 2,999 dams from Alpine and Saanen breeds. The aim of this study was to investigate methods of genomic evaluation including αs1 casein information to improve accuracies. Genomic evaluations were investigated both for the whole population (multi-breed pooling together Alpine and Saanen animals) and also for each breed separately. Several genomic models including 50k BeadChip genotypes and/or αs1 casein information were tested on protein content. A classical ssGBLUP approach using only SNP information was compared to (1) weighted ssGBLUP approaches allocating a weight to each SNP and used to construct the genomic relationship matrix (G), (2) a TABLUP approach based on some SNP selected to construct G and (3) a gene content approach using both αs1 casein and SNP genotypes in a multiple trait model. Cross validation analyses consisted into splitting the population of the 905 males genotyped with the goat 50K BeadChip into a training set of 554 males born before 2007 and at test set of 351 young males born between 2008 and 2012. Results on validation correlations were slightly better in multi-breed than within breed for all approaches. Selecting the most associated SNP to protein content with TABLUP approach did not improve validation correlations compared to ssGBLUP. Results of gene content approach and weighted ssGBLUP approaches were close and showed an improvement of the validation correlations (1% to 4.58% in multi-breed analyses, 0.4% to 3.10% in Alpine population and 2.90% to 5.15% in Saanen population) compared to ssGBLUP approach.