Optimization of wheat breeding programs using an evolutionary algorithm achieves enhanced genetic gain through strategic resource allocation

Abstract

Plant breeding is a complex process that involves trade-offs among competing breeding objectives and limited resources. Despite the necessity for optimization of breeding program design, the inherent complexity can make optimization challenging. Most often, a predefined set of scenarios is compared or a single parameter of a breeding scheme is assessed in depth. In a previous study, we developed an optimization pipeline, utilizing stochastic simulations and an evolutionary algorithm, suitable for the joint optimization of multiple class and continuous parameters. Here, we assess the applicability of our framework to realistic plant breeding schemes. For this, a wheat line breeding scheme simulated using AlphaSimR and a wheat hybrid breeding scheme simulated using MoBPS were considered. Both schemes were further optimized using our optimization pipeline. When aiming to maximize genetic gain while maintaining a fixed budget, the breeding program design suggested by our optimization pipeline results in 32.6% higher genetic gain compared to a proposed baseline wheat line breeding program. When adjusting the breeding objective to put 20% of the weight on the maintenance of genetic diversity, genetic gain still increased by 4.5% while maintaining 9.1% higher genetic diversity compared to the baseline. Similarly, 4.6%/8.8% higher genetic gains for the male/female part of the hybrid breeding scheme compared to its baseline scenario were obtained. Results highlight the importance of optimizing breeding program design to improve breeding efficiency, with the suggested pipeline offering breeders a powerful framework to refine breeding designs, balance breeding goals, and enhance competitiveness, profitability, and sustainability. Core Ideas Evolutionary algorithm can improve breeding design by optimizing many breeding decisions simultaneously Our framework is compatible with various backend simulators, enabling broad application across platforms Optimized designs enhance genetic gain and diversity, demonstrating greater overall breeding program efficiency Genetic gain in wheat line program increased by over 30% compared to a baseline program without additional costs Plain Language Summary Modern plant breeding programs must make many complex decisions, such as how many plants to test or where to grow them while staying with tight budgets. These choices are often connected and involve trade-offs between goals like improving genetic gain, maintaining diversity, and reducing costs. In this study, we used an evolutionary algorithm framework to evaluate thousands of breeding program designs and identify the most effective ones. The optimized programs outperformed the respective baseline programs, increasing genetic gain by up to 32% and preserving 9% more genetic diversity - all without additional costs. This flexible framework can support a wide range of breeding decisions and adapt to different program types, providing breeders with a powerful tool to improve outcomes and use resources more efficiently.