Using a single-step SNP BLUP maternal-effect model for calving trait genomic evaluation in German Holsteins

Abstract

A multi-parity maternal-effect linear animal model has been used for conventional evaluation of calving traits in Germany. Bull EBV of direct and maternal effects of calving ease and stillbirth are submitted to Interbull bull MACE evaluation that uses a single-trait single-effect model. For a single-step evaluation of the calving traits, we applied a single-step SNP BLUP model to the national calving data integrated with foreign pseudo-phenotype data of bulls. We used phenotype, genotype and pedigree data from German official genomic evaluation in August 2021. About 25 million national calving records and deregressed EBV of c.a. 113,000 bulls with foreign data were evaluated. The number of calving cows or calves with national calving data exceeded 31 millions. A total of 1,003,041 genotyped animals were considered in the single-step evaluation without any approximation of genomic information. A genomic validation was performed for the single-step and the current multi-step model. SNP effects from the single-step model were shown to have less bias and were more correlated between the truncated and the full evaluation than those from the multi-step model. Using genomic EBV of validation bulls from the full and truncated evaluations, we conducted a linear regression validation test and obtained less satisfactory validation results for all the calving traits than the test-day or conformation traits. More inflation was seen in the direct than the maternal genetic effects, in particular the stillbirth direct effect with largest overestimation. The overestimation seemed to be more evident for the younger than the older validation bulls. Increasing the residual polygenic variance improved little in reducing the prediction inflation. A special bull reference population for the calving traits failed to markedly reduce the overestimation either. However, removing genotype data of older bulls, born before 2005, resulted in favorable validation results, though to a limited degree. Post-processing genomic EBV of young candidates seemed to be unavoidable for a routine implementation of the single-step model for the calving traits.