On the occasion of the PhD defence of Yvonne Wientjes, on Friday January 22 the Animal Breeding and Genomics Centre organized a Breed4Food seminar entitled “Genomic prediction using multiple breeds”. Presentations were given by Dr Hans Daetwyler (Australia), Prof Mogens Lund (Denmark) and Prof Theo Meuwissen (Norway). In total 65 people attended the seminar and several viewers watched the streaming of the seminar via WURTV. The main conclusion was that utilizing biological information may improve genomic selection, albeit that there is no consensus yet on the optimal approach to achieve this.
The central theme of the seminar was the use of multiple breeds in genomic prediction. Dr Hans Daetwyler started with presenting that genomic prediction using sequence data holds the promise to improve persistency of genomic predictions across generations and breeds, because the sequence data potentially includes all causal variants. Still, it is very important to be able to distinguish between causal and non-causal variants. Using biological information, either from annotation of the variants or gene expression profiles, was shown to be a potentially interesting approach. Prof Mogens Lund took a different approach, and selected sequence variants based on associations detected in the data, either within or across breeds. After strict selection of such associated variants, adding them to commonly used 50k SNP chips improved genomic prediction reliabilities by up to 6 percentage points. It is, however, important to only use those sequence SNPs that are very close to the causal variants. Prof Theo Meuwissen presented a method that can increase genomic prediction reliability by using information of both genotyped and ungenotyped animals, that is more sophisticated than the commonly used single step GBLUP. The presented method was better able to impute the genotypes for the ungenotyped animals, and yielded more accurate and less biased predictions.
These excellent presentations on how to utilise whole genomic sequence information in genomic prediction, together with available biological information, are closely related to the current research within Breed4Food.
The seminar can still be viewed through WUR-TV.