CBGG

Introducing genomic prediction approaches at an early stage (i.e., selecting the best crosses) is less disruptive than at advanced stages (identifying the best progeny) in terms of the breeding process and resources involved. Here, we tried to assess the reliability of a predictive approach in an applied breeding context. First, we developed a genomic selection model to estimate trait values and validated it on existing progenies. It was then used to predict the mean (µ) and genetic variance (VG) for each cross among a large set of simulated progenies. The degree of agreement between predicted and observed values for key traits indicated that the predictive model provided an adequate degree of accuracy. Crosses predicted to be superior produced progeny that persisted longer in the breeding process suggesting that the predictions are consistent with reality. The predicted correlations between traits known to be correlated (e.g., DON-GYD) were concordant with observed and expected correlations signifying that the properties of these simulated progeny were in line with expectations. Among the 30,000 potential crosses that could be made between lines comprising the training population, only 2.2% were predicted to exhibit a low correlation between DON and GYD and just 0.13% were predicted to produce progeny in which the top lines could combine high GYD with reduced DON. Even in the absence of empirical proof that genomic prediction can outperform classical practice, the results obtained here appear encouraging regarding the potential of such an approach in barley breeding programs.

genomic prediction; performing progeny; promising crosses; genetic variance; quantitative traits; Fusarium resistance; yield; barley; breeding program