Utilization Of Genotyping-by-Sequencing for Genomic Selection in the Kansas State University Wheat (Triticum aestivum) Breeding Program

Genomic selection (GS) is a breeding method that uses dense, genome-wide molecular markers to predict the breeding value of individuals. We have developed and validated a training population for the Kansas State University (KSU) winter wheat (Triticum aestivum) breeding program.  Genotyping-by-sequencing (GBS) was used to generate a whole-genome profile of 832 lines from the KSU Hays wheat breeding program. Lines were phenotyped at one to three dryland locations with three replications for agronomic traits including yield and best linear unbiased predictors for each line were calculated. TASSEL, a bioinformatics pipeline, was used to order and internally align the reads to discover and genotype 60,844 SNPs. A principal component analysis demonstrated a distinct lack of population structure in the breeding program. Ridge regression was used for model validation. Model efficiency varied widely based on the population subset. Subsets by year, alone, yielded correlations from -0.67 to 0.64, but most of these were not significantly different from zero (p > 0.05). When using a training population of lines from 1999-2008 to make predictions on lines from 2009-2011, however, a correlation of 0.258 was calculated (p-value = 0.0197). In 2013, GBS and GS will be used in combination with phenotypic selection of lines to give recommendations on selection.