Application Of Low Density Panels In Genomic Selection On Meat Quality Traits In Simmental Cattle

Genome-wide association analysis and genomic selection were widely used in animal breeding. This study performed GWAS on carcass and meat quality trait using BovineHD, predicted genomic breeding value and estimated accuracy using low density panels on fatty acid component traits in Simmental.Using Compressed mixed linear model(CMLM) and linear model(LM), it performed GWAS on knuckle, hind shank and bone weight, detected 186 significant SNPs(P<10-5), of which 55 SNPs identified by both methods. The majority of SNPs harbored on LAP3, LCORL, FAM184 B on BTA6 and PLAG1, TOX and XKR4 on BTA14, overlapping with previously reported carcass weight and bone weight QTLs.It performed GWAS on six meat quality traits, which contained marbling score, fat color, total fatty acid(TFA), saturated fatty acid(SFA), monounsaturated fatty acid(MUFA) and polyunsaturated fatty acid(PUFA), detected 91 significant SNPs of which 44 SNPs were identified via two approaches. Markers which multiply associated with three fatty acid component traits were adjacent to MYC on BTA14. Most of significant SNP located on QTLs related to marbling score, 12 th fat thickness and fatty acid components.It applied different low density panels including evenly spaced panels(ELD) and selected marker panels(SLD) to predict genomic breeding value on four fatty acid component traits, evaluating accuracy via cross-validation. Selected marker methods were on the basis of the absolute value of marker effect that was estimated by Bayes A and BayesB methods or significant value. Evenly spaced low density panels integrated BovineHD with known low density panels, contained 3000 to 40000 markers. Accuracy of GEBV using 9K panel was high in ELD, but was lower than using SLD that suggested by several simulations in genomic prediction. The number of selected markers on the basis of effect and significant value were 300 to 30000. Using 7K panel esitimated GEBV was promising. Among three methods, BayesB-based was better than others, BayesA and significant had a slight difference. Cross-validation using IBS cluster was higher accuracy than random group.