Identification of Predictor Genes for Feed Efficiency in Beef Cattle by Applying Machine Learning Methods to...

Area of Science:

• Genomics
• Bioinformatics
• Animal Science

Background:

• Machine learning (ML) methods show promise in identifying genes from large transcriptome datasets.
• Previous studies have not compared combined ML methods for predicting high feed efficiency (HFE) and low feed efficiency (LFE) animals.

Purpose of the Study:

• To evaluate and compare the performance of different ML methods, individually and in combination, for classifying feed efficiency (FE) in animals.
• To identify a minimal set of genes that accurately predict FE status using RNA sequencing data.

Main Methods:

• Utilized RNA sequencing data from five tissues of HFE and LFE Nellore bulls.
• Compared five analytical methods: t-test, edgeR, Random Forests (RFs), Extreme Gradient Boosting (XGBoost), and a combined RF and XGBoost (RX) approach.
• Assessed gene subset utility for classification using Support Vector Machine (SVM) and performed gene co-expression network analysis.

Main Results:

• The combined RX method identified the smallest gene subset (117 genes) with the highest classification accuracy, outperforming individual methods.
• Gene co-expression network analysis supported the biological relevance and interactivity of ML-identified genes.
• The study demonstrated the effectiveness of ML combinations for identifying predictive genes from transcriptome data.

Conclusions:

• A combination of ML methods (RX) offers a powerful approach for identifying key genes related to feed efficiency from large transcriptome datasets.
• This strategy significantly improves the accuracy of classifying animals based on feed efficiency.
• The findings highlight the potential for ML in advancing genetic selection for improved animal production traits.