Application of a Weighted Genomic Relationship Matrix to Breeding Value Prediction for Egg Production in Laying Hens

In pedigree based genetic evaluations distantly related individuals provide little or no information to the estimated breeding values of selection candidates. With genomic prediction, the effects of markers are typically assumed to remain constant across generations such that close and distant relatives contribute equally to estimates of effects. In the current study we propose to weight the difference between pedigree and genomic relationships depending on the pedigree based distance between individuals. If the pedigree based relationship (A) was greater than 0.45, the genomic relationship was used, otherwise relationships were shrunk towards pedigree by reducing the difference between A and G by n times A, where n was equal to 1, 2 or 3. The egg production rate (%) of non-genotyped progeny of genotyped parents where incorporated in the analysis by fitting 0.5 of the parental breeding values and effects for Mendelian sampling.  Genomic relationships based on 42k SNPs were constructed among 3,908 genotyped animals with up to 13,000 phenotypes (individual and progeny).  The accuracies of prediction for 303 animals in validation (progeny of the last generation of training) were 0.13, 0.17, 0.18 and 0.19 for the standard G matrix and weighted G matrices with n = 1, 2 and 3 respectively. Thus, reducing the weighting on genomic relationships for more distant relatives increased accuracies of prediction for egg production in the studied population.