Optimizing QTL introgression via stochastic simulations: an example of the IRRI rice breeding program

Abstract A key limitation in the ability of breeding programs to leverage benefits of major-gene marker-assisted selection is the availability of those genes in appropriate elite germplasm. In this context, our study compared three strategies to develop new recipients for QTL introgression (Background recovery (BG), Selective sweep (SS), and Breeding values (BV)) in a short-term breeding program (over five breeding cycles). Furthermore, we evaluated two different numbers of recipients (10 and 20) in the introgression process and how they influence the population performance and the QTL fixation over cycles. Finally, we used rice as a model of a self-pollinated crop and implemented stochastic simulations. Each strategy was simulated and replicated 40 times. Regardless of the selection strategy used, the QTL introgression resulted in substantial penalties in yield performance. However, introducing fewer new parents to the augmentation process minimized this effect. Conversely, the time required to achieve fixation of target QTLs showed substantial differences, with selection for BV during augmentation out-performing other methods. Overall, the BV_10 strategy (10 parents selected based on genomic estimated breeding values) displayed the best trade-off between reduced penalty from introducing new QTLs with a reasonable speed at which those QTLs can achieve fixation over subsequent breeding cycles.