Optimization of recurrent rapid cycle breeding in maize for sustained long-term genetic improvement via stochastic simulations.

In recent years, the turnover of germplasm in plant breeding has substantially increased as the use of genomic information allows for earlier selection and the integration of controlled growing environments reduces the time to reach a particular growing stage. However, high generation turnover and intensive selection of lines before own yield trials are performed come at the risk of a drastic reduction of genetic diversity and lower prediction accuracies. To this end, we investigate strategies to cope with these challenges in a maize rapid cycle breeding scheme using stochastic simulations employing the software MoBPS. We find that genetic gains soon reach a plateau when only the original breeding material is phenotyped. Updating the training data set via additional phenotyping of crosses or doubled haploid lines ensures long-term progress with a gain of 6.80/6.95 genetic standard deviations (gSD) for the performance as a cross/DH after 30 cycles of breeding compared with 3.40/4.28 without additional phenotyping. Introducing genetic material from outside the breeding pool to introduce novel genetic diversity led to a further increase to 9.34/7.89 gSD. In particular, for the management of genetic diversity, further modifications of breeding program design are analysed to optimize the number of selected lines per cycle and to account for the relatedness of F2 plants in the selection using the software AlphaMate. Balancing short-term genetic gains with long-term diversity preservation is crucial for sustainable breeding. MoBPS provides a tool for quantifying these effects and provides solutions specific to the respective breeding program.