CBGG

Background: Relatively little effort has been made yet to optimize the allocation of resources when using genomic predictions to maximize the return to investment in terms of genetic gain per unit of time and cost.Methods: We built a simulation pipeline in the R environment designed to become a decision tool to help breeders adjusting breeding schemes, according to their either short or long-term objectives. We used it to explore different scenarios in order to investigate under which conditions (at what step of the breeding program) genomic predictions could improve genetic gain. For a given budget per cycle, we compared 36 scenarios, varying strategies (phenotypic selection PS or genomic selection + phenotypic selection: GPS), trait heritability, relative selection rate at two key steps and genotyping cost. With GPS strategy, we also optimized mating using genomic predictions. The reference population is a 20 years historical data set from the INRAE-Agri-Obtentions bread wheat breeding program. We simulated 3 cycles of 5 years selection.Results: We showed that GPS selection using mating optimization significantly improved genetic gain for all scenarios while GPS without mating optimization and PS had similar efficiency in terms of genetic gain. Our results also highlighted that the loss of genetic diversity over successive cycles was faster using GPS strategies. Those were more efficient to increase favourable allele frequency, rare alleles in particular.

breeding scheme; genomic prediction; mating optimization; economic constraint; bread wheat