Optimizing self-pollinated crop breeding employing genomic selection: from schemes to updating training sets
Abstract Long-term breeding schemes employing genomic selection (GS) can boost the response to selection per year. Although several studies show that GS delivers a higher response to selection, only a few analyze the best strategy to employ it, specifically how often and in what manner the training set (TS) should be updated. Therefore, we used stochastic simulation to compare in a long-term breeding program of a hypothetical self-pollinated crop five different strategies to implement GS in the line fixation stage and four methods and sizes to update the TS. Moreover, among breeding schemes, we proposed a new approach for using GS to select the best individuals in each F2 progeny based on genomic estimated breeding and divergence and crossed them to generate a new recombination event. Finally, we compared these schemes to the traditional phenotypic selection and drift. Our results showed that using GS in F2 followed by the phenotypic selection of new parentals in F4 was the best scenario. Furthermore, adding a new set of training data every cycle (over 800) to update the TS maintains the accuracy at satisfactory levels for many more generations, showing that more data is better than optimizing the genetic relationship between TS and the targeted population in a closed system. Hence, we believe that these results may help breeders optimize GS in their programs and improve genetic gain in long-term schemes.