Genomic prediction of avian influenza infection outcome in layer chickens

Abstract

Avian influenza (AI) is a devastating poultry disease that currently can be controlled only by liquidation of affected flocks. In spite of typically very high mortality rates, a group of survivors was identified and genotyped on a 600K single nucleotide polymorphism (SNP) chip to identify genetic differences between survivors, and age-and genetics-matched controls from unaffected flocks. In a previous analysis of this dataset, a heritable component was identified and several regions that are associated with outcome of the infection were localized but none with a large effect. For complex traits that are determined by many genes, genomic prediction models using all SNPs across the genome simultaneously are expected to optimally exploit genomic information. In this study, we evaluated the diagnostic value of genomic estimated breeding values for predicting AI infection outcome within and across two highly pathogenic avian influenza viral strains and two genetic lines of layer chickens using receiver operating curves. We show that genomic prediction based on the 600K SNP chip has the potential to predict disease outcome especially within the same strain of virus (area under receiver operating curve above 0.7), but did not predict well across genetic varieties (area under receiver operating curve of 0.43).