Developing tools for efficient genomic evaluation in crossbreeding

Published on
March 22, 2017

Genomic selection is becoming the standard approach for improving quantitative traits in livestock. To predict genomic breeding values, the Breed4Food companies Cobb, Hendrix Genetics, and Topigs Norsvin use the so-called single-step genomic Best Linear Unbiased Prediction (ssGBLUP). The ssGBLUP is a method that uses a combination of the genomic and pedigree relationship to estimate genomic breeding values. With the increasing amount of genotypes, this method will requires too much time and memory soon. Current research at Wageningen University & Research, Animal Breeding and Genomics, focuses on the feasibility of the ssGBLUP in the context of crossbreeding schemes, and on the development of tools for its routine implementation.

The single-step genomic evaluation, or ssGBLUP, is appealing due to its simultaneous use of data from genotyped and non-genotyped animals through the combination of the genomic and pedigree relationship matrices. These matrices are used to account for genetic relationships between individuals when estimating genomic breeding values. They are also the origin of two main issues of the ssGBLUP. First, ssGBLUP could become infeasible with more than 150,000 genotyped animals. Indeed, the computation of the inverse of the genomic relationship matrix requires large amounts of memory and time. Second, the combination of genomic and pedigree relationship matrices requires that they are compatible. This compatibility is not straightforward, especially when genotyped purebred and crossbred animals are involved in the same genomic evaluation. Within the research program “Bright ideas for innovations in animal breeding?”, researchers at Wageningen University & Research, Animal Breeding and Genomics, aimed to tackle both issues. Solutions they are working on, are  the implementation of an algorithm for a fast and efficient inversion of the genomic relationship matrix (so-called “Algorithm for Proven and Young animals”; APY), and therecent concept of metafounders in the software calc_grm for routine implementations. A metafounder is a pseudo-individual that describes the relationships among the founders of the pedigree. The implementation of both concepts is performed in collaboration with the development of the software for genetic and genomic evaluations, MiXBLUP2.0.

Currently, the APY and use of metafounders are implemented in the software calc_grm, and are under testing. The APY approach was tested on data simulated to mimic a crossbreeding scheme. The results showed that APY achieved similar accuracies as the full genomic relationship matrix. Several recommendations for an optimal implementation of the APY in the context of crossbreeding schemes have been proposed to the industrial partners. Second, the method of metafounders also implemented in calc_grm was tested on real data of purebred and crossbred layers, provided by the Breed4Food partner Hendrix Genetics. Results showed that the method of metafounders allowed to adjust the pedigree relationships to be compatible with genomic relationships.

Following this project, the Breed4Food partners are jointly testing the tools on field datasets similar to those used for their routine evaluations and exchanging their experiences.