Assessment of the potential for genomic selection to improve resistance to fusarium stalk rot in maize

Fusarium stalk rot (FSR), caused by <i>Fusarium verticilliodes</i>, is a serious disease in maize. Resistance to FSR is complexly inherited. Thus, an investigation was carried out to predict and validate the genomic estimated breeding values (GEBVs) for FSR resistance. Three doubled haploid (DH) populations induced from F₁ and F₂ of the cross VL1043 × CM212 and F₂ of the cross VL121096 × CM202 were used in the current study. Six different parametric models (Genomic-Best Linear Unbiased Predictors (GBLUP), BayesA, BayesB, BayesC, Bayesian least absolute shrinkage and selection operator (BLASSO), and Bayesian Ridge Regression (BRR)) were employed to estimate the prediction accuracy. Further, the accuracy of predicted genomic estimated breeding value (GEBV) for FSR resistance was assessed using five-fold cross-validation and independent validation. The training population (TP) size and marker density were optimized by considering different proportions of training set (TS) and validation set (VS) and varying marker density from 40 to 100%. The estimates of descriptive statistics and genetic variability parameters, which include mean, standardized range, genetic variance, phenotypic and genotypic coefficients of variations, broad sense heritability, and genetic advance as per cent mean (GAM), were relatively higher in DH F₂s than those in DH F₁s. Prediction accuracies displayed an increasing trend with an increase in the proportion of training set size and marker density in all three DH populations. The TS:VS proportion of 75:25 in DH F₁ (VL1043 × CM212) and DH F₂ (VL121096 × CM202), and 80:20 in DH F₂ of VL1043 × CM212 resulted in greater prediction accuracy than other TS:VS proportions. Study of linkage disequilibrium (LD) decay pattern across all the populations indicated that the number of markers employed were sufficient to conduct a genomic prediction (GP) study in two DH F₂ populations of crosses VL1043 × CM212 and VL121096 × CM202. Prediction accuracies of 0.24 and 0.17 were recorded for FSR resistance in independent validation when DH F₂ of cross VL121096 × CM202 was used for validation and DH F₁ and DH F₂s from the cross VL1043 × CM212 as training sets. A significant positive correlation of FSR resistance between the DHs selected based on their GEBVs and those selected based on test cross performance indicated the efficiency of genomic prediction models.