A novel hyper-parameter can increase the prediction accuracy in a single-step genetic evaluation

Abstract

AbstractThe H-matrix best linear unbiased prediction (HBLUP) method has been widely used in livestock breeding programs. It can integrate all information, including pedigree, genotypes, and phenotypes on both genotyped and non-genotyped individuals into one single evaluation that can provide reliable predictions of breeding values. The existing HBLUP method (e.g., that implemented in BLUPf90 software) requires hyper-parameters that should be adequately optimised as otherwise the genomic prediction accuracy may decrease. In this study, we assess the performance of HBLUP using various hyper-parameters such as blending, tuning and scale factor in simulated as well as real data on Hanwoo cattle. In both simulated and cattle data, we show that blending is not necessary, indicating that the prediction accuracy decreases when using a blending hyper-parameter < 1. The tuning process (adjusting genomic relationships accounting for base allele frequencies) improves prediction accuracy in the simulated data, confirming previous studies, although the improvement is not statistically significant in the Hanwoo cattle data. We also demonstrate that a scale factor, α, which determines the relationship between allele frequency and per-allele effect size, can improve the HBLUP accuracy in both simulated and real data. Our findings suggest that an optimal scale factor should be considered to increase the prediction accuracy, in addition to blending and tuning processes, when using HBLUP.Author SummaryDespite significant advancements in genotyping technologies, the capability to predict the phenotypes of complex traits is still limited. H-matrix best linear unbiased prediction (HBLUP) method has been used to tackle this limitation to demonstrate a promising prediction accuracy. However, the performance of HBLUP depends heavily on the optimisation of hyper-parameters (e.g. blending and tuning). In this study, we introduce a scale factor (α), as a new hyper-parameter in HBLUP, which accounts for the relationship between allele frequency and per-allele effect size. Using simulation and real data analysis, we investigate the impact of the hyper-parameters (blending, tuning, and scale factor) on the performance of HBLUP. In general, the blending process may not improve the prediction accuracy for simulation and cattle data although a marginally improved prediction accuracy is observed with a blending hyper-parameter = 0.86 for one of carcass traits in the cattle data. In contrast, the tuning process can increase the HBLUP accuracy particularly in simulated data. Furthermore, we observe that an optimal scale factor plays a significant role in improving the prediction accuracy in both simulated and real data, and the improvement is relatively large compared with blending and tuning processes. In this context, we propose considering the scale factor as a hyper-parameter to increase the predictive performance of HBLUP.