Two simple methods to improve the accuracy of the genomic selection methodology.

Area of Science:

• Agricultural Science
• Genetics
• Biotechnology

Background:

• Genomic selection (GS) is a powerful tool in plant and animal breeding.
• Practical implementation of GS faces challenges due to various influencing factors.
• Conventional GS, formulated as a regression problem, often shows low sensitivity in identifying elite individuals.

Purpose of the Study:

• To enhance the prediction accuracy of genomic selection.
• To develop novel methods for more effective selection of top candidate individuals.
• To improve the practical applicability of genomic selection in breeding programs.

Main Methods:

• Reformulating genomic selection from a regression problem to a binary classification problem.
• Implementing a postprocessing step to adjust thresholds for classification.
• Training binary classification models using a defined threshold for top vs. non-top lines.

Main Results:

• Both proposed methods significantly outperformed the conventional regression model in prediction accuracy.
• The postprocessing method demonstrated superior performance compared to the reformulation method.
• Sensitivity, F1 score, and Kappa coefficient showed substantial improvements with the proposed methods.

Conclusions:

• A simple postprocessing method effectively improves genomic regression models without reformulation.
• The proposed methods significantly enhance the selection of top candidate lines.
• Both methods are easily adoptable for practical breeding programs, ensuring improved selection outcomes.