Use of Cluster Analysis for Identifying Metafounders

Area of Science:

• Animal breeding and genetics
• Quantitative genetics
• Genomic evaluation

Background:

• The metafounder approach uses a relationship matrix (Γ) to integrate pedigree founder relationships into the numerator relationship matrix (A), creating A^Γ.
• Traditional metafounder definitions (e.g., by country, sex, breed) can limit their ability to accurately represent population structure, potentially causing inconsistencies in Single-Step Models when combining pedigree (A) and genomic (G) relationship matrices into H.
• Genotypes offer reliable information on true population structure.

Purpose of the Study:

• To investigate an approach for transferring population structure information from genotyped animals to ungenotyped ancestors using metafounders.
• To determine the optimal number of metafounders (nMF) by harmonizing inbreeding trends in G and A^Γ and monitoring Γ.
• To evaluate the impact of genotype-informed metafounders on prediction quality in Single-Step genomic evaluations under different genetic trend scenarios.

Main Methods:

• An unsupervised clustering approach was used to assign pedigree founders to metafounders.
• Single-Step genomic evaluations were performed with varying numbers of metafounders (nMF).
• A semi-stochastic simulation based on Fleckvieh genotypes was used to assess prediction quality for traits with and without genetic trends.

Main Results:

• Modeling metafounders defined by population structure slightly reduced prediction quality for traits with no genetic trend, but remained stable around the optimal nMF.
• For traits with a strong genetic trend, prediction quality was improved compared to a standard Single-Step model.
• The largest improvements in prediction quality were observed within the range of the proposed optimal nMF.

Conclusions:

• Defining metafounders using population structure derived from genotypes can enhance genomic evaluation accuracy, particularly for traits with strong genetic trends.
• The proposed method for determining the optimal number of metafounders provides a stable and effective range for improving prediction quality.
• This genotype-informed metafounder approach offers a valuable refinement for Single-Step genomic evaluations by better capturing underlying population structure.