Sunday 02 February 2025
Scientists have long been searching for ways to optimize breeding programs, ensuring that they produce the best possible offspring while minimizing risks and uncertainties. A team of researchers has made a breakthrough in this area by developing two new methods for robust optimal contribution selection (OCS), which takes into account uncertainty in breeding values.
Traditional OCS methods assume that breeding values are known exactly, but in reality, there is always some degree of uncertainty involved. This can lead to suboptimal decisions and reduced genetic gain. The new methods, developed by the researchers, use concepts from robust optimization to address this issue.
The first method involves reformulating the OCS problem as a bilevel optimization problem, where the inner problem minimizes the difference between the true breeding value and its estimated counterpart, subject to uncertainty constraints. This approach is particularly useful when dealing with small cohorts or limited data.
The second method uses sequential quadratic programming (SQP) to solve the robust OCS problem. SQP is a powerful optimization technique that can handle complex problems by iteratively solving smaller sub-problems. In this case, the algorithm starts with an initial solution and refines it by adding new constraints based on the uncertainty in breeding values.
To test these methods, the researchers used simulated data from a breeding program and compared their performance to traditional OCS methods. The results showed that the robust OCS methods outperformed traditional approaches, particularly when dealing with larger cohorts or more complex problems.
One of the key findings was that the HiGHS solver performed better than Gurobi for solving the robust OCS problem using SQP. This is likely due to the fact that HiGHS is designed specifically for convex optimization problems, which are well-suited for robust OCS.
The researchers also developed a Python package called robustocs, which allows users to implement these methods and solve robust OCS problems using either HiGHS or Gurobi solvers. This package provides an easy-to-use interface for breeding program managers and geneticists to optimize their selection decisions while accounting for uncertainty in breeding values.
Overall, this research represents a significant step forward in the field of breeding program optimization. By incorporating uncertainty into the decision-making process, these methods can help ensure that breeding programs are more resilient and effective in producing high-quality offspring.
Cite this article: “Optimizing Breeding Programs with Uncertainty in Mind”, The Science Archive, 2025.
Breeding Programs, Optimal Contribution Selection, Robust Optimization, Uncertainty, Breeding Values, Genetic Gain, Bilevel Optimization, Sequential Quadratic Programming, Sqp, Highs Solver, Gurobi Solvers
