Transition of the UK dairy national evaluation to across-breed and single-step genomic evaluation: somatic cell counts as a case trait

Abstract

Current UK genomic evolutions follow a two-step approach; initially, a genetic evaluation based solely on pedigree information, followed by a Single Nucleotide Polymorphism Best Linear Unbiased Prediction (SNPBLUP) analysis for genotyped animals using de-regressed proofs, including MACE proofs from Interbull. Nowadays, with recent advances in computational feasibility and growing interest in across-breed genomic evaluations for dairy herds, there is a compelling need to adopt a single-step across-breed genomic evaluation approach within the UK dairy industry. The single-step method offers notable advantages by simultaneously incorporating genotypes and both recent and historical pedigree and phenotypic data into a single analysis. This integration enhances the accuracy of genetic predictions across diverse breeds, accelerates genetic progress, and improves selection efficiency. This study aims to evaluate the impact of using genomic information and compare the prediction ability of single-step genomic evaluation (using ssSNPBLUP method) and pedigree-based genetic evaluation (PedBLUP) employing cross-validation techniques (Linear Regression method). The trait analysed was somatic cell count (SCC), using data from the UK national evaluations as of the December 2024 official run. The dataset included 11,271,959 animals in the pedigree and 19,056,954 SCC records from 7,527,712 cows. Foreign information was incorporated for 182,844 bulls, with adjustments made to avoid double counting of domestic data. Genotypic data was available for 891,480 animals, imputed at 79,051 SNPs using findhap.f90 V3. Analyses were performed using the MiX99 V23.1026 software, applying an ssSNPBLUP model with 10% polygenic effects. The validation group comprises bulls born after 2016 and cows born after 2018, whose records are set to missing. Results showed a genomic accuracy improvement of up to 54% in cows when comparing ssSNPBLUP to PedBLUP. Among bulls, the greatest gain was observed in Holsteins (+33%), followed by Guernsey and Ayrshire (+30%), and Jersey (+20%). Level bias and dispersion bias was slightly reduced in ssSNPBLUP relative to PedBLUP. Overall, the findings demonstrate that single-step genomic evaluation is a promising and efficient approach for enhancing prediction accuracy in UK dairy cattle.