Bayesian genomic prediction of junctional epidermolysis bullosa in sheep

Authors: BURAK KARACAÖREN

Abstract: Junctional epidermolysis bullosa (JEP) is a heritable skin and mucosa disorder in association with mendelian mutations in sheep. The purpose of this investigation is to explore the relationship between different priors, linkage disequilibrium, and single nucleotide polymorphism (SNP) selection methods and accuracy of Bayesian GP of JEP in sheep. Ninety-two Spanish Churra sheep breed genotyped by 40668 SNP markers. Bayes C? was shown to have slightly higher predicted accuracy [0.724 (0.113)] by unselected data. Prediction performance of the Bayesian GP models was found to be similar after correction for LD. There was a significant difference between predicted accuracies due to the SNP selection by ranked p values of whole and training only dataset using linear model. The relevance of genetic architecture in conjugate to the prior distributions was clearly supported by the unselected data. The most obvious finding of this study is that preselection of SNPs referring to genetic architecture of the phenotype may lower the needs of computational load.

Keywords: Bayesian models, genomic prediction, junctional epidermolysis bullosa

Full Text: PDF