Integrating Bioinformatics and Machine Learning for Genomic Prediction in Chickens

Area of Science:

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

Background:

• Genomic prediction is vital for animal breeding, but classical models struggle with complex genetic data.
• Novel approaches are needed to improve predictive accuracy for multiple traits.

Purpose of the Study:

• To evaluate machine learning (ML) methods for genomic prediction in Rhode Island Red chickens.
• To compare ML performance against traditional methods like rrBLUP and BayesA.
• To assess the impact of incorporating GWAS-identified SNPs into ML models.

Main Methods:

• Genotyping 4190 chickens using an Illumina 50K SNP chip.
• Applying ML algorithms and classical bioinformatics methods to predict 10 economic traits.
• Comparing predictive accuracy using Pearson correlation and RMSE against rrBLUP and BayesA.

Main Results:

• ML algorithms outperformed rrBLUP and BayesA for body weight and eggshell strength prediction.
• rrBLUP and BayesA showed higher accuracy for egg number prediction (2-58%).
• Incorporating GWAS SNPs into ML models increased predictive accuracy by 0.1-27% across most traits.

Conclusions:

• ML methods offer a powerful alternative for genomic prediction in poultry.
• Combining GWAS with ML enhances predictive accuracy for economic traits.
• Integrated approaches hold significant potential for future animal breeding strategies.