Optimization of breeding program design through stochastic simulation with evolutionary algorithms.

Area of Science:

• Animal Breeding and Genetics
• Quantitative Genetics
• Computational Biology

Background:

• Effective resource allocation is critical for modern breeding program success.
• Previous methods using kernel regression for optimization required extensive simulations, limiting effectiveness with many parameters.
• Balancing genetic gain, diversity, and costs necessitates careful assessment of breeding program design parameters.

Purpose of the Study:

• To develop a more effective and general optimization framework for breeding programs.
• To improve the efficiency of resource allocation and parameter optimization in breeding schemes.
• To reduce the computational burden associated with optimizing complex breeding programs.

Main Methods:

• Proposed an optimization framework combining kernel regression concepts with an evolutionary algorithm.
• Utilized stochastic simulations to evaluate the performance of potential breeding program parameter settings.
• Implemented the evolutionary algorithm within a Snakemake workflow for scalable distributed computing.

Main Results:

• The evolutionary algorithm achieved optimization with a massively reduced number of simulations compared to previous methods.
• The new framework demonstrated improved computing time and scalability when incorporating class variables and more parameters.
• The algorithm stabilized around the same optimum, indicating robust performance.

Conclusions:

• The proposed evolutionary algorithm framework offers a more efficient and scalable approach to optimizing breeding programs.
• This method effectively balances trade-offs between breeding goals and costs, leading to improved genetic gain and diversity.
• The framework's ability to handle more parameters and class variables enhances its applicability to complex breeding schemes.