robustocs: Robust optimal contribution selection

Area of Science:

• Animal breeding and genetics
• Quantitative genetics
• Optimization methods

Background:

• Traditional selective breeding methods like Optimal Contribution Selection (OCS) and Truncation Selection (TS) do not account for uncertainty in breeding value estimates.
• This uncertainty can impact the short-term competitiveness and long-term sustainability of breeding programs.

Purpose of the Study:

• To introduce Robust Optimal Contribution Selection (ROCS) by incorporating robust optimization concepts.
• To develop and implement computational solutions for the ROCS problem.
• To compare the performance of ROCS against traditional methods.

Main Methods:

• Formulation of the Robust Optimal Contribution Selection (ROCS) problem using robust optimization.
• Development of two solution approaches: conic optimization and sequential quadratic programming.
• Implementation of these methods in the Python package 'robustocs', utilizing Gurobi and HiGHS solvers.

Main Results:

• Favorable performance observed for ROCS solved using sequential quadratic programming with the HiGHS solver.
• Classical TS and OCS identified as special cases of the robust selection formulations (RTS and ROCS).
• ROCS and RTS reduce uncertainty in genetic gain and coancestry at a cost of slightly reduced genetic gain compared to TS and OCS.

Conclusions:

• ROCS provides a robust framework for selective breeding that manages uncertainty in genetic gain.
• The 'robustocs' package offers practical tools for implementing advanced selection strategies.
• ROCS enhances the long-term sustainability of breeding programs by balancing genetic gain with reduced uncertainty.