Study of the Effect of Haplotype Block Length on Improving the Accuracy of Genomic Prediction Using Bayesian Methods in Sheep

Linkage disequilibrium (LD) advancement map and the specification of population-level haplotype block structures are parameters that are helpful for managing the study of the Genome wide Association (GWAS), and to comprehend the nature of non-linear relationship among phenotypes and genotype. Compared with single nucleotide polymorphisms (SNP), genomic prediction fitting haplotype alleles and improve prediction accuracy; but the increase in accuracy belong how the Haplotype block are characterized. The aim of this study was to test the optimal size for haplotype length in genomic predictions.

The Haplotype alleles were defined according the SNP alleles in not covering blocks 125 Kb, 250 Kb, 500 Kb, and 1 Mb. The Haplotype alleles with frequencies below 1, 2.5, 5 or 10% are eliminated. Two methods, Bayes A and Bayes B, were used to predict the genomic effects of SNPs and haplotypes. From Bayes A and B methods to predict the genomic effects of SNPs and haplotypes in three traits with three levels of heritability (milk production (h2 = 0.1), carcass weight (h2 = 0.3) and body weight in Maturity (h2 = 0.45) was used.

The highest genomic prediction obtained in body weight at maturity by Bayesian method B (0.652) during 250 kb haplotypic block and the lowest by Bayesian method A in milk production (0.407) during haplotypic block 1 Mb. Haplotype blocks of 250 kb with a frequency threshold of 1% provided the highest genomic prediction accuracy. Comparing Bayes A and Bayes B methods, Bayes B method provided higher estimation accuracy in both SNP-based and haplotype allele-based models.

: Placing haplotype alleles instead of SNPs in the statistical model, if the haplotype length is properly defined, improves the accuracy of genomic prediction.