Approximating genomic reliabilities for national genomic evaluation

Authors: LIU, ZENGTING - Collaborator; Vanraden, Paul; LIDAUER, MARTIN - Natural Resources Institute Finland (LUKE); CALUS, MARIO - Wageningen University And Research Center; BENHAJALI, HAIFA - Swedish University; JORJANI, HOSSEIN - Swedish University; DUCROCQ, VINCENT - Institut National De La Recherche Agronomique (INRA)

Submitted to: Interbull Annual Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 11/30/2017
Publication Date: 11/30/2017
Citation: Liu, Z., Van Raden, P.M., Lidauer, M., Calus, M., Benhajali, H., Jorjani, H., Ducrocq, V. 2017. Approximating genomic reliabilities for national genomic evaluation. Interbull Bulletin. 51:75-85.

Interpretive Summary: Many approximate formulas have been used to report the reliability or accuracy of genomic predictions. More precise, standard methods were developed by an Interbull working group. These should efficiently provide reliability estimates that are more directly comparable across countries. The methods can adjust for many differences in the data available within and among the populations and traits. Validation of reliability estimates should give animal breeders more confidence that genomic predictions are as accurate as claimed.

Technical Abstract: With the introduction of standard methods for approximating effective daughter/data contribution by Interbull in 2001, conventional EDC or reliabilities contributed by daughter phenotypes are directly comparable across countries and used in routine conventional evaluations. In order to make published genomic reliabilities comparable across countries and consistent with conventional reliabilities, a working group for genomic reliability calculation developed a new method that is feasible for any number of genotyped animals and also adjusts theoretical model genomic reliabilities based on genomic validation results. The first step of the proposed reliability method calculates reliabilities contributed by SNP genotypes via an efficient software snp_blup_rel. This new genomic reliability method accounts for the residual polygenic effect in genomic evaluation and is applicable to both single-step and multi-step genomic models. The adjustment procedure makes the changes in genomic reliabilities reflecting the changes in GEBV and ensures candidates genomic reliabilities from an early evaluation being consistent with later genomic reliabilities when the animals have received phenotype data. The proposed reliability method was applied to a large German Holstein population. Adjustment factors for the theoretical model genomic reliabilities were derived based on a genomic validation study via Interbull GEBV Test. Results from the test implementation for German Holsteins demonstrated high efficiency and feasibility of the proposed genomic reliability method. Several aspects have been discussed for future optimisations. All involved countries were requested to test the software snp_blup_rel and proposed genomic reliability method. Depending on the country feedbacks, the software and the proposed genomic reliability method will be fine-tuned towards an official implementation by all the involved countries.