Preselection statistics and Random Forest classification identify population informative single nucleotide...

Area of Science:

• Animal Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Commercial SNP arrays offer tools for breed differentiation.
• Statistical methods are needed to identify informative SNPs from large datasets.

Purpose of the Study:

• To compare SNP preselection methods for identifying discriminating markers in dairy cattle.
• To evaluate the effectiveness of 48-SNP and 96-SNP panels for breed assignment.
• To assess the utility of Random Forest classification combined with preselection techniques.

Main Methods:

• Compared Delta, Fst, and Principal Component Analyses (PCA) for SNP preselection.
• Utilized Random Forest classification on SNP data from six dairy cattle breeds.
• Created and evaluated 48-SNP and 96-SNP panels based on discrimination ability and linkage disequilibrium.

Main Results:

• 96-SNP panels generally showed better breed discrimination than 48-SNP panels.
• The PCA-chrom derived 96-SNP panel achieved the lowest error rates, especially for minor breeds.
• Selected SNPs were located near genes influencing phenotypic and production traits.

Conclusions:

• Random Forest combined with preselection techniques effectively identifies population-informative SNPs.
• The PCA-chrom method provides a valuable panel for differentiating minor and admixed cattle breeds.
• SNP panels can be tailored to identify markers associated with specific breed traits.