Genome-enabled prediction of reproductive traits in Nellore cattle using parametric models and machine learning...

Area of Science:

• Animal Genetics
• Genomic Prediction
• Machine Learning in Animal Breeding

Background:

• Genomic prediction is crucial for improving reproductive traits in Nellore cattle.
• Evaluating advanced machine learning methods alongside traditional models is essential for optimizing breeding strategies.

Purpose of the Study:

• To assess the predictive performance of Support Vector Regression (SVR), Bayesian Regularized Artificial Neural Network (BRANN), and Random Forest (RF) for genomic prediction of reproductive traits.
• To compare these machine learning models against parametric models like Genomic Best Linear Unbiased Predictor (GBLUP) and Bayesian Least Absolute Shrinkage and Selection Operator (BLASSO).

Main Methods:

• Utilized a 5-fold cross-validation strategy to evaluate predictive ability.
• Computed average prediction accuracy (ACC) and mean squared errors (MSE) for traits including age at first calving (AFC), scrotal circumference (SC), early pregnancy (EP), and stayability (STAY).
• Defined ACC based on trait type (continuous or categorical) to ensure appropriate performance measurement.

Main Results:

• Support Vector Regression (SVR) demonstrated slightly higher prediction accuracies across all reproductive traits compared to GBLUP and BLASSO.
• SVR improved prediction accuracy for AFC by 6.3% (vs. GBLUP) and 4.8% (vs. BLASSO), for SC by 8.3%, for EP by 4.5%, and for STAY by 4.8%.
• Random Forest (RF) and Bayesian Regularized Artificial Neural Network (BRANN) did not show competitive predictive ability compared to the parametric models.

Conclusions:

• Support Vector Regression (SVR) is a suitable machine learning method for genome-enabled prediction of reproductive traits in Nellore cattle.
• The optimal kernel bandwidth parameter for SVR is trait-dependent, necessitating fine-tuning during the training phase for maximum efficacy.
• Machine learning models, particularly SVR, offer potential advancements in genomic selection for cattle breeding programs.