Predicting genetic merit in Harnali sheep using machine learning techniques.

Area of Science:

• Animal Science
• Genetics
• Machine Learning

Background:

• Genomic and phenotypic data are crucial for improving animal breeding.
• Accurate prediction of genetic merits is essential for efficient livestock improvement.

Purpose of the Study:

• To evaluate seven machine learning algorithms for predicting the genetic merits of Harnali sheep.
• To identify the most effective algorithm for optimizing breeding programs in this breed.

Main Methods:

• Seven machine learning algorithms (KNN, MLR, BR, SVM, ANN, RF, GBM) were tested.
• Pedigree and phenotypic data from 2036 Harnali lambs (1998-2021) were used.
• Models were trained on 75% of data and tested on 25%, evaluated using R², RMSE, MAE, AIC, BIC, and bias.

Main Results:

• Gradient Boosting Machine (GBM) was the top-performing model.
• GBM achieved a coefficient of determination (R²) of 0.64 and predictive accuracy (r) of 0.80.
• GBM demonstrated superior performance with lower error and bias metrics compared to other algorithms.

Conclusions:

• Machine learning, particularly GBM, shows significant potential for enhancing Harnali sheep breeding programs.
• Accurate prediction of breeding values using GBM can accelerate genetic progress in sheep populations.
• The study highlights the utility of advanced computational methods in modern animal genetics and breeding.