Assessing genomic selection prediction accuracy in a dynamic barley breeding population and comparing gain between genomic and phenotypic selection in barley

Abstract

Prediction accuracy of genomic selection has been previously evaluated through simulation and cross-validation; however validation based on progeny performance in a plant breeding program has not been investigated thoroughly. We evaluated several prediction models in a dynamic barley breeding population comprised of 647 six-row lines using four traits differing in genetic architecture and 1,536 SNP markers. The breeding lines were divided into six sets designated as one parent set and five consecutive progeny sets comprised of representative samples of breeding lines over a five-year period. We used these data sets to investigate the effect of model and training population composition on prediction accuracy over time. We found little difference in prediction accuracy among the models confirming prior studies that found the simplest model, RR-BLUP, to be accurate across a range of situations. In general, we found that using the parent set was sufficient to predict progeny sets with little to no gain in accuracy from generating larger training populations by combining the parent set with subsequent progeny sets. The prediction accuracy ranged from 0.03 to 0.99 across the four traits and five progeny sets. We explored characteristics of the training and validation populations (marker allele frequency, population structure, and linkage disequilibrium) as well as characteristics of the trait (genetic architecture and heritability). Fixation of markers associated with a trait over time was most clearly associated with reduced prediction accuracy for the mycotoxin trait DON. Higher trait heritability in the training population and simpler trait architecture were associated with greater prediction accuracy.Genomic selection is a marker based selection method that promises to improve and accelerate the breeding process in plants and animals. Numerous studies have investigated the gain per unit time; however very limited studies have directly compared gains from genomic and phenotypic selection using empirical data. In this study, we used five consecutive sets of breeding lines to compare the gain between genomic and phenotypic selection. In each set, about ninety six barley breeding lines were phenotypically evaluated for yield, FHB resistance, and DON accumulation. A set 168 historic parental lines were used as a training population to predict the performance of the selection candidate sets using RR-BLUP. All lines were genotyped using 1,536 SNP markers (BOPA1) for all seven barley chromosomes. The best performing 10% of the breeding lines in each year were selected using the two schemes and revaluated together in several trials in Minnesota and North Dakota. We made direct comparison between genomic and phenotypic selection in two selection candidate sets for yield and five sets for FHB resistance and DON accumulation. We assessed the relative efficiency of genomic over phenotypic selection and changes in the genetic similarity using the two selection schemes. Results showed similar response to selection between genomic and phenotypic selection in most cases. Genomic selection resulted in more genetic similarity only for FHB resistance; however, for yield and DON concentration no changes in the genetic similarity were detected between genomic and phenotypic selection. In addition, we assessed the use of phenotypic selection for FHB, genomic selection for FHB, and genomic selection for DON as indirect selection methods to select for low DON concentration. We did not find significant differences between direct and indirect selection methods.