An optimized single-step SNP BLUP model for calf fitness in German Holstein

Abstract

A single-step SNP BLUP was introduced in routine official evaluation of German Holstein in April 2025 for all traits, including an early-measured trait, calf fitness, defined as calf survival between day 3 and 15 months after birth. Prior to the single-step model implementation, a mixed reference population of bulls and calves was set up for the calf fitness genomic evaluation using a multi-step genomic model. During the testing phase of the single-step model, an unrealistic, strong genetic trend of calf fitness was observed in genotyped animals, when compared to the multi-step genomic model or pedigree-based conventional model. Having searched for plausible causes for the overestimation, we detected a much lower mortality rate for genotyped calves than non-genotyped ones, particularly for the early periods from day 3 to 120 after birth. Although all female calves were genotyped under the whole-herd genotyping scheme in Germany, farmers did not always take genotyping sample right after the birth of a calf, causing a delay in genotyping for the early periods of the trait calf fitness. In addition, there were limited economic incentives for farmers to genotype dead calves. To solve the overprediction bias of the calf fitness evaluation, we developed a new single-step model by using only genotypes of sires of all female calves with phenotypic data. Genomic breeding values of the genotyped calves and all other genotyped animals were indirectly predicted based on SNP effect estimates and residual polygenic effect estimates of all the genotyped sires from the new single-step model. Genomic validation showed a slightly higher accuracy of the new single-step model using sire genotypes than the original model using genotypes of all animals. In comparison to a significant over-prediction for the original model, the new single-step model using only the sire genotypic records gave an almost unbiased genomic prediction. Genetic trends in genotyped AI bulls or female animals were no longer overestimated with the new single-step model. The problem of inflated genomic prediction of the original single-step model seems to be solved by using only the genotypic data of sires of female calves.