Genomic Selection for Cashmere Traits in Inner Mongolian Cashmere Goats Using Random Forest, Gradient Boosting...

Area of Science:

• Animal Genomics
• Machine Learning Applications
• Quantitative Genetics

Background:

• Genomic prediction is crucial for livestock breeding.
• Machine learning (ML) offers advanced methods for analyzing high-dimensional genomic data.
• Traditional Genomic Selection (GS) methods can be enhanced by ML algorithms for improved accuracy and efficiency.

Purpose of the Study:

• To identify the optimal machine learning algorithm for genome-wide selection of cashmere traits in Inner Mongolian cashmere goats.
• To compare the predictive accuracy of four ML algorithms: Random Forest (RF), Extreme Gradient Boosting Tree (XGBoost), Gradient Boosting Decision Tree (GBDT), and LightGBM.
• To evaluate the impact of hyperparameter optimization on ML model performance for cashmere trait prediction.

Main Methods:

• Genomic prediction accuracy was assessed using genotype and phenotype data from 2299 Inner Mongolian cashmere goats.
• Four ML algorithms (RF, XGBoost, GBDT, LightGBM) were employed and compared.
• Parameter optimization was performed for each ML algorithm to enhance prediction accuracy.

Main Results:

• LightGBM achieved the highest accuracy for fiber length (56.4%), RF for cashmere production (35.2%), and GBDT for cashmere diameter (40.4%).
• ML methods improved accuracy by 0.8-2.7% compared to GBLUP.
• Parameter optimization led to average accuracy improvements of 2.9% (cashmere fineness), 2.7% (cashmere yield), and 3.8% (fiber length).
• XGBoost showed the lowest prediction accuracy across the studied traits.

Conclusions:

• Machine learning algorithms, particularly LightGBM, RF, and GBDT, demonstrate significant potential for enhancing genomic selection of cashmere traits.
• Hyperparameter tuning is essential for maximizing the predictive accuracy of ML models in genomic prediction.
• ML approaches offer a valuable alternative to traditional methods, improving computational efficiency and prediction accuracy in goat breeding programs.