Effects of truncation and false positives in selection of markers for genomic prediction

Abstract

Association studies are often used to select genetic markers for genomic prediction, requiring truncation with a balance between power and false positives. Using simulation, the aim for this study is to test the effects of truncation on prediction accuracy and particularly the impact of false positives in SNP selection. Bonferroni and Benjamini-Hochberg False Discovery Rate methods were tested. Our study suggested that except for polygenic traits, truncation with a more lenient threshold such as the Benjamini-Hochberg False Discovery Rate increases the genomic prediction accuracy. In an inbred population, false positives could contribute positively to the accuracy especially for a oligogenic trait, although further study would be needed to generalize this result. Our study suggested that for a polygenic trait, all markers should be included in genomic prediction, and if SNP selection were to be applied a lenient threshold for truncation would be desirable.