Utilizing Multivariate Genomic Prediction to Predict Wheat Spike Characteristics in Soft Red Winter Wheat (Triticum aestivum L.)

Genomic prediction (GP) can increase genetic gain by allowing for the selection of traits earlier in the breeding cycle. Spike morphology traits are of interest because of their relationship with grain yield. To better understand the prediction capabilities of agronomic traits and spike architecture traits, a population of 594 soft red winter wheat inbred lines was evaluated for traits including heading date, grain weight spike −1 , thousand kernel weight, and kernel spike −1 and spike architecture traits spikelets spike −1 (SPS), spike length (SL), and spike width (SW). Univariate GP models were utilized for each spike trait, and multivariate GP models were used to predict spike architecture traits using each combination of all seven traits for a total of 63 multivariate models per spike trait. Significantly higher cross‐validation mean prediction accuracies ( p ( Z ) < 0.05) were observed for 52 models for SPS, 0 models for SW, and 13 models for SL compared with the univariate models of each trait. Methods in this study could be used for GP of spike architecture traits with further analysis of forward validation accuracy.