Prediction of Hanwoo Cattle Phenotypes from Genotypes Using Machine Learning Methods.

Area of Science:

• Animal Breeding and Genetics
• Genomics
• Machine Learning in Agriculture

Background:

• Hanwoo cattle are economically vital in Korea, with increasing demand for meat production.
• Genomic selection aims to enhance genetic progress in Hanwoo breeding programs.
• Improved statistical methods are needed to boost genomic prediction accuracy.

Purpose of the Study:

• To compare the predictive performance of Random Forest (RF), Extreme Gradient Boosting (XGB), and Support Vector Machine (SVM) against GBLUP.
• To evaluate machine learning methods for predicting key carcass traits in Hanwoo cattle.
• To identify optimal methods for genomic breeding value prediction in Hanwoo.

Main Methods:

• Utilized phenotypic and genotypic data (53,866 SNPs) from 7324 Hanwoo cattle.
• Applied machine learning algorithms: Random Forest (RF), Extreme Gradient Boosting (XGB), and Support Vector Machine (SVM).
• Compared predictive performance with the traditional GBLUP method for carcass weight (CWT), marbling score (MS), backfat thickness (BFT), and eye muscle area (EMA).

Main Results:

• XGBoost achieved the highest predictive correlation for CWT and MS.
• GBLUP demonstrated the best predictive correlation for BFT and EMA.
• No machine learning method significantly outperformed GBLUP in terms of mean squared error of prediction.

Conclusions:

• While XGBoost shows promise for specific carcass traits, GBLUP remains the recommended method for predicting genomic breeding values in Hanwoo cattle.
• The study highlights the ongoing need for advanced statistical approaches in livestock breeding.
• GBLUP provides reliable predictions for essential carcass traits in Hanwoo.