Validation of the Prediction Accuracy for 13 Traits in Chinese Simmental Beef Cattle Using a Preselected Low-Density SNP Panel.

Ling Xu,Lupei Zhang,Huijiang Gao,Bo Zhu,Lingyang Xu,Wentao Cai,Xue Gao,Junya Li,Lei Xu,Qunhao Niu,Hongwei Li,Yan Chen,Zezhao Wang

doi:10.3390/ani11071890

Abstract

Simple SummaryTo reduce the breeding costs and promote the application of genomic selection (GS) in Chinese Simmental beef cattle, we developed a customized low-density single-nucleotide polymorphism (SNP) panel consisting of 30,684 SNPs. When comparing the predictive performance of the low-density SNP panel to that of the BovineHD Beadchip for 13 traits, we found that this ~30 K panel achieved moderate to high prediction accuracies for most traits, while reducing the prediction accuracies of six traits by 0.04–0.09 and decreasing the prediction accuracy of one trait by 0.2. For the remaining six traits, the usage of the low-density SNP panel was associated with a slight increase in prediction accuracy. Our studies suggested that the low-density SNP panel (~30 K) is a feasible and promising tool for cost-effective genomic prediction in Chinese Simmental beef cattle, which may provide breeding organizations with a cheaper option and greater returns on investment.Chinese Simmental beef cattle play a key role in the Chinese beef industry due to their great adaptability and marketability. To achieve efficient genetic gain at a low breeding cost, it is crucial to develop a customized cost-effective low-density SNP panel for this cattle population. Thirteen growth, carcass, and meat quality traits and a BovineHD Beadchip genotyping of 1346 individuals were used to select trait-associated variants and variants contributing to great genetic variance. In addition, highly informative SNPs with high MAF in each 500 kb sliding window and in each genic region were also included separately. A low-density SNP panel consisting of 30,684 SNPs was developed, with an imputation accuracy of 97.4% when imputed to the 770 K level. Among 13 traits, the average prediction accuracy levels evaluated by genomic best linear unbiased prediction (GBLUP) and BayesA/B/Cπ were 0.22–0.47 and 0.18–0.60 for the ~30 K array and BovineHD Beadchip, respectively. Generally, the predictive performance of the ~30 K array was trait-dependent, with reduced prediction accuracies for seven traits. While differences in terms of prediction accuracy were observed among the 13 traits, the low-density SNP panel achieved moderate to high accuracies for most of the traits and even improved the accuracies for some traits.

Highlights

Genomic prediction (GP), which uses genome-wide markers to predict direct genomic estimated breeding values (DGVs), has been widely studied in breeding programs for plants [1,2,3] and domestic animals [4,5,6]
The linkage disequilibrium (LD) of this population was quantified via the r2 value with the 770 K chip after quality control
We found that the LD decreased from 0.61 to 0.01 within the 2 Mb window (Figure 1)

Summary

Introduction

Genomic prediction (GP), which uses genome-wide markers to predict direct genomic estimated breeding values (DGVs), has been widely studied in breeding programs for plants [1,2,3] and domestic animals [4,5,6]. In real-world contexts, the true QTLs are unknown, and markers must be selected as proxies to explain genetic variance [7,8] In this setting, increasing the available markers could allow all QTLs in LD to be linked with at least one marker, which in turn would be beneficial for the GP. Previous studies have demonstrated that the usage of high-density markers (~770 K) can contribute to an improvement of prediction accuracy in comparison with moderate- (~50 K) or low-density (~30 K) markers [5,9,13,14]. This increase is limited compared with the high cost of genotyping.

Objectives

Methods

Results

Discussion

Conclusion