Genomic Selection Strategies to Accelerate Breeding Progress in Wheat.

Genomic selection (GS) has successfully been used in plant breeding to reduce breeding cost and time while improving selection efficiency. However, the most limiting aspect of genomic selection for plant breeding programs is its actual implementation, especially figuring out strategies to select superior parents, to handle genotype by environment interaction (GxE), and to choose individuals to include in the training population. The objective of this work was to present different strategies for implementing genomic selection in breeding programs. First, we compared strategies to exploit GxE in GS using mixed models. Second, we evaluated strategies for selecting the best crosses (i.e. crosses that have the highest chance of producing superior progeny). Finally, we compared strategies for optimizing the training population to predict specific sets of individuals. We used two large wheat (Triticum aestivum L.) populations from the INIA and CIMMYT breeding programs. A set of advanced wheat lines evaluated for yield in 35 location-year combinations were used to compare strategies for GxE obtaining either overall or by-environment predictions for different sets of environments. Higher predictive ability was obtained by characterizing and modeling GxE in the GS context. Strategies to predict best crosses using combinations of mid-parent value and variance prediction accounting for linkage disequilibrium (V_LD) or assuming linkage equilibrium (V_LE) were compared in the INIA and the CIMMYT breeding programs for grain yield, grain protein content, mixing time, and loaf volume. The mid-parent values of the crosses were the most important factor for progeny performance for grain yield. However, in the case of quality traits, the variance of the cross played an important role. We compared strategies to evaluate optimize the training population for a specific population. These strategies will improve GS implementation in breeding programs.