Estimation of individual animal SNP-BLUP reliability using full Monte Carlo sampling

Area of Science:

• Quantitative Genetics
• Animal Breeding
• Genomics

Background:

• SNP-BLUP models estimate genomic breeding values, but require computationally intensive inversion of mixed model equations (MME).
• The inclusion of residual polygenic (RPG) effects increases MME size, exacerbating computational challenges.

Purpose of the Study:

• To introduce and evaluate a full Monte Carlo (MC) sampling-based method for approximating reliability in SNP-BLUP.
• To compare the performance of the MC approximation method against the traditional genomic BLUP (GBLUP) model.

Main Methods:

• Developed a full Monte Carlo (MC) sampling approach to approximate reliability in SNP-BLUP.
• Evaluated the MC method's performance on two large datasets with varying numbers of genotyped animals and SNP markers.
• Compared computational demands and accuracy against the GBLUP model.

Main Results:

• The MC approximation method demonstrated significantly lower computational demands compared to standard MME inversion.
• A tendency for the MC method to overestimate reliability was observed, particularly for animals with inherently low reliability and high RPG effect weights.
• Increasing the number of MC samples effectively reduced the overestimation of reliability.

Conclusions:

• The MC approximation offers a computationally efficient alternative for estimating reliability in SNP-BLUP models.
• Careful consideration of the number of MC samples is necessary to mitigate potential overestimation of reliability, especially in specific population structures.
• The MC method provides a viable approach for large-scale genomic evaluations where computational efficiency is paramount.