Figure 1.

Schematic representation of the breeding program applied in this study. Two synthetic populations $Sy{n}_{N_R}^{\left(1\right)}$ and $Sy{n}_{N_R}^{\left(2\right)}$ were separately created by using $N_R$ recombination generations from $N_p$ parental gametes drawn from one ancestral population [with short- (SR) or long-range linkage disequilibrium (LR)]. If the training set (TS) and the recurrent selection candidates (RSC) were related, TS and RSC were sampled from the same synthetic $Sy{n}_{N_R}^{\left(1\right)},$ and if they were unrelated, they were drawn from separate synthetics $Sy{n}_{N_R}^{\left(1\right)}$ and $Sy{n}_{N_R}^{\left(2\right)}.$ In each cycle of recurrent selection, $N_s=10$ individuals were selected and recombined to establish the next generation.