Assessing the Impact of Different Mixing Strategies on Genomic Prediction Accuracy for Beef Cattle Breeding Values in Multi-Breed Genomic Prediction.

Although genomic selection can accelerate livestock breeding, its application in many countries is hindered due to the limited size of reference populations. To address this issue, researchers have explored methods of combining multiple breeds to create reference populations, aiming to enhance the accuracy of genomic prediction. The main objective of this study was to evaluate the impact of the construction of mixed reference populations at different genetic distance levels on the accuracy of multi-breed genome prediction in multi-breed beef cattle populations using three evaluation methods: GBLUP, ssGBLUP, and wGBLUP. In order to study the effect of genetic correlation on multiple populations and to resolve the optimal mixing ratio, we considered six scenarios, including (1) population A as the main body, where the nearest 10% of individuals in populations B and C were added; (2) population A was the main body, where the 15% of individuals with the closest genetic distance in groups B and C were added; and (3) population A as the main body, where the 20% of individuals in populations B and C with the closest genetic distance were added. Our results suggest that the wGBLUP model can be enhanced when the mixing ratio is 15%, and the wGBLUP model shows higher accuracy in predicting populations with different LD decay patterns. Among them, whether combined with PopB or PopC, the wGBLUP model shows better prediction ability than the GBLUP and ssGBLUP models. However, when the mixing ratio is 10% or 20%, the accuracy of the three models is less than 15%, and the wGBLUP and ssGBLUP models show high and stable accuracy. Our results highlight the importance of considering the proportion of mixing between different populations when using genetic assessment models to predict accuracy, especially for endemic beef cattle breeds with different genetic structures and LD patterns and limited resources. However, this study also has some limitations. First, the determination of the optimal mixing ratio still needs further exploration, especially for populations with different genetic structures and LD patterns. Second, future studies can introduce more advanced models to further improve prediction accuracy.