Identification of Target Chicken Populations by Machine Learning Models Using the Minimum Number of SNPs

Area of Science:

• Genetics
• Animal Breeding
• Bioinformatics

Background:

• Consumer confidence in chicken origin is crucial for commercial value and protecting native genetic resources.
• Accurate classification of chicken populations is essential for market differentiation and conservation efforts.

Purpose of the Study:

• To identify an optimal, minimal set of genetic markers for classifying specific chicken populations.
• To evaluate the efficacy of machine learning algorithms and genomic analyses in marker selection for population classification.

Main Methods:

• Utilized a 600k high-density single nucleotide polymorphism (SNP) array on 283 samples from 20 chicken lines.
• Applied machine learning (AdaBoost, Random Forest, Decision Tree), genome-wide association study (GWAS), linkage disequilibrium (LD) analysis, and principal component analysis (PCA) for marker selection.
• Selected 96 LD-pruned SNPs as the best combination and identified minimum marker sets (8-36 SNPs) using machine learning models.

Main Results:

• Achieved high classification accuracy rates (up to 99.6%) with minimal marker sets (8-36 SNPs).
• Selected marker combinations increased genetic distance and fixation index (Fst) between populations.
• A verification study with additional breeds confirmed the distinct classification capability of the selected markers.

Conclusions:

• A small set of selected SNPs can efficiently and accurately classify chicken populations.
• The integrated approach of GWAS, PCA, and machine learning is effective for identifying optimal marker combinations.
• This method supports the commercial value of specific chicken populations and the conservation of native genetic resources.