Multiple-breed genomic evaluations by using a reduced pool of SNP-markers

Tuesday, July 22, 2014
Exhibit Hall AB (Kansas City Convention Center)
Massimo Cellesi , UniversitÓ di Sassari, Sassari, Italy
Nicol˛ P.P. Macciotta , UniversitÓ di Sassari, Sassari, Italy
Paolo Ajmone-Marsan , UniversitÓ Cattolica del Sacro Cuore, Piacenza, Italy
Attilio Rossoni , Associazione Nazionale Allevatori Razza Bruna, Bussolengo, Italy
Gabriele Marras , UniversitÓ di Sassari, Sassari, Italy
Giustino Gaspa , UniversitÓ di Sassari, Sassari, Italy
Corrado Dimauro , UniversitÓ di Sassari, Sassari, Italy
Abstract Text:

Large reference populations (RP) of genotyped and phenotyped individuals are required to obtain reliable predictions in genomic selection programs. For small breeds, however, assembling such RPs could result particularly challenging. In this study, a multibreed approach was used to enhance the size of the RP. Data were genotypes of 2,054 Italian Holstein and 634 Brown Swiss bulls, respectively, genotyped with the Illumina’s 50K BeadChip. Phenotypes were deregressed proofs (DRP) for milk, fat and protein yield. An empirical technique, named Maximum Difference Analysis (MDA), was used to select a restricted pool of SNP-markers significantly associated with the considered trait (T). In each breed, animals were ranked according to a T. The best (B) 10% and the worst (W) 10% individuals were selected and the genotypic frequencies were evaluated. For each SNP, the maximum genotypic frequency in B and, in correspondence, the frequency for the same genotype in W were recorded and the difference between the two frequencies was calculated. A bootstrap procedure was then implemented to derive a posterior probability distribution that was used to declare a SNP positively associated with T. Markers negatively associated with T were detected through the same procedure with the only difference that, for each SNP, the maximum genotypic frequency was recorded in W. A BLUP model was used to estimate marker effects that were then used to calculate genomic breeding values of Brown Swiss younger bulls (50animals). Three datasets were used: all original SNPs, only the MDA selected markers (MDA_SNP) and the MDA_SNP for Brown Swiss plus MDA_SNP obtained for Holstein. MDA selected, for both breeds, around 1,500 markers for each trait. Accuracies of genomic predictions (table 1) evaluated by using MDA_SNPs in the multi-breed scenario were greater than values obtained with all markers and in the single-breed scenario. Results suggested that the MDA applied to a small genotyped bovine population increases accuracies of genomic predictions of about 10%. A further improvement can be obtained in a multibreed scenario.

Table. Accuracies of genomic predictions for Brown Swiss.





Brown Swiss all SNP




Brown Swiss MDA_SNP




Holstein + Brown Swiss MDA_SNP




Keywords: SNP reduction, multiple breeds, genomic selection