Solving computational challenges in single-step genomic predictions

Offentlig prosjektsammendrag

At Geno, we currently use a single-step genomic best linear unbiased prediction (ssGBLUP) approach to estimate breeding values for the Norwegian Red dairy cattle breed. This method leverages the inverse of the relationship matrix, incorporating both pedigree and genomic information, to predict breeding values. Since its implementation in Geno breeding program in 2016, this approach has tripled the speed of genetic progress for the Norwegian Red dairy cattle breed. Each year, approximately 30,000 additional animals are genotyped, leading to a current total of about 240,000 genotyped animals. As a result, calculating the inverse of the relationship matrix has become computationally demanding. It is therefore crucial to find an alternative method that maintains the genetic progress achieved with ssGBLUP while reducing computational demands. Several alternative methods have been developed and described in recent literature, but these require thorough testing on real breeding program data. We will specifically test these methods on the Geno breeding program. The results of this project will provide critical insights into the differences between these alternative methods when applied to our breeding program. Ultimately, this knowledge will enable us to implement the best alternative to the current ssGBLUP approach in the routine evaluation of breeding values for the Norwegian Red dairy cattle breed.