Training population selection and use of fixed effects to optimize genomic predictions in a historical USA winter wheat panel

Abstract

Key message The optimization of training populations and the use of diagnostic markers as fixed effects increase the predictive ability of genomic prediction models in a cooperative wheat breeding panel. Abstract Plant breeding programs often have access to a large amount of historical data that is highly unbalanced, particularly across years. This study examined approaches to utilize these data sets as training populations to integrate genomic selection into existing pipelines. We used cross-validation to evaluate predictive ability in an unbalanced data set of 467 winter wheat ( Triticum aestivum L.) genotypes evaluated in the Gulf Atlantic Wheat Nursery from 2008 to 2016. We evaluated the impact of different training population sizes and training population selection methods (Random, Clustering, PEVmean and PEVmean1) on predictive ability. We also evaluated inclusion of markers associated with major genes as fixed effects in prediction models for heading date, plant height, and resistance to powdery mildew (caused by Blumeria graminis f. sp. tritici) . Increases in predictive ability as the size of the training population increased were more evident for Random and Clustering training population selection methods than for PEVmean and PEVmean1. The selection methods based on minimization of the prediction error variance (PEV) outperformed the Random and Clustering methods across all the population sizes. Major genes added as fixed effects always improved model predictive ability, with the greatest gains coming from combinations of multiple genes. Maximum predictabilities among all prediction methods were 0.64 for grain yield, 0.56 for test weight, 0.71 for heading date, 0.73 for plant height, and 0.60 for powdery mildew resistance. Our results demonstrate the utility of combining unbalanced phenotypic records with genome-wide SNP marker data for predicting the performance of untested genotypes. Electronic supplementary material The online version of this article (10.1007/s00122-019-03276-6) contains supplementary material, which is available to authorized users.

Genomic Selection at Preliminary Yield Trial Stage: Training Population Design to Predict Untested Lines

“…Overall, results from our study were encouraging regarding the prediction of preliminary lines, even for highly polygenic and complex traits like yield and test weight. The model's predictive ability, with cross validation, showed moderate to high accuracies for the different traits, in agreement with previous studies on wheat [11,13,15,[26][27][28].…”

“…Isidro et al [13] applying different strategies to define the TP, obtained for wheat populations with a mild population structure, prediction accuracies ranging from 0.2-0.5, noting an increase of prediction accuracy with increases in TP size. Recently, Sarinelli et al [15] obtained prediction accuracies ranging from 0.4 to 0.6 under different methods of selecting the TP and different TP sizes.…”

“…Another key consideration of genomic selection is the design of the training population (TP). How to design the TP has been extensively discussed in the literature, with a primary focus on size of the TP [6][7][8][9][10][11]; and how to optimize and update the training population [12][13][14][15]. Another factor that influences the accuracy of genomic predictions is the relatedness between the TP and the validating population (VP) [10,13,16,17].…”

“…Therefore, when designing the training population, many questions arise as to whether wheat panels with related and unrelated material, advanced lines from the breeding programs or full sibs of the same families may be the best choice. Some authors have based GS studies on wheat panels, regional trials involving hundreds of lines, or sets of advanced lines in the breeding program [1,11,13,15,19,20]. All of these studies in wheat obtain GEBVs for the different traits by applying a cross validation approach.…”

Genomic Prediction and Indirect Selection for Grain Yield in US Pacific Northwest Winter Wheat Using Spectral Reflectance Indices from High-Throughput Phenotyping

“…Galiano-Carneiro et al [38] previously observed improved accuracy for Fusarium head blight-related traits in winter triticale (× Tritosecale) through fitting significant markers with >5% genetic variance as fixed effect in a weighted ridge regression model. Maximum gains in accuracy were also observed when combinations of multiple height and phenology genes were included as fixed effects in a prediction model for yield in a historical US winter-wheat panel [39].…”