Definition of metafounders based on population structure analysis

Area of Science:

• Animal Genetics
• Quantitative Genetics
• Bioinformatics

Background:

• Conventional numerator relationship matrix (A) formulation has limitations in stratified breeding populations.
• Combining A with genomic relationship matrix (G) in single-step evaluations can introduce bias.
• Need to identify and incorporate population stratification for accurate genetic predictions.

Purpose of the Study:

• To identify population stratification using genomic data (ADMIXTURE).
• To integrate stratification information into matrix A (generating A_meta).
• To improve the compatibility between A and G matrices.

Main Methods:

• Iterative approach using ADMIXTURE software to detect 2-40 strata.
• Metafounder methodology to incorporate strata into matrix A, creating A_meta.
• Regression analysis and comparison of matrix properties (mean, diagonal) to evaluate compatibility.
• Tested on 85,249 Brown Swiss animals' genotype and pedigree data (S1 and S2 datasets).

Main Results:

• Regression of A on G showed poor fit (intercept -0.489, slope 0.780) in the initial setup.
• Integrating stratification (k=7 for S1) improved regression (intercept -0.028, slope 1.087).
• Optimal stratification (k=24 for S2) resulted in negligible differences in matrix means and diagonals (intercept -0.020, slope 0.998).

Conclusions:

• Integrating stratification information into matrix A significantly improves A and G compatibility.
• This approach enhances the accuracy of single-step genetic evaluations in stratified populations.
• Balancing data is crucial for population structure analysis in dairy breeding with substructures.