Marker effect p-values for single-step SNP-BLUP genomic models
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Single-step Single Nucleotide Polymorphism best linear unbiased prediction (ssSNPBLUP) is a comprehensive method for obtaining genomically enhanced breeding values for animals and SNP effects in a single evaluation. The ssSNPBLUP model integrates phenotypic, pedigree, and genomic data for genomic evaluations. However, there has been no framework for estimating the reliability and p-values of the SNP effects obtained from a ssSNPBLUP genomic model. This study investigates the reliability and significance of the SNP effects estimated using a ssSNPBLUP framework in German Limousin (LIM) and Holstein (HOL) cattle populations.
Methods
This study introduces a novel approach for calculating p-values within the ssSNPBLUP framework and compares it to a conventional single-marker regression GWAS approach. SNP reliabilities were computed using prediction error variances of SNP effect estimates, enabling the identification of statistically significant SNP markers. LIM data included weaning weight (200-DW) evaluated with a maternal effect BLUP model, while HOL data comprised production traits (milk yield, protein yield, fat yield, and somatic cell score) analysed via a random regression test-day model.
Results
The results reveal significant SNP effects in both LIM and HOL evaluations, with notable differences attributed to the size of the reference populations. Average SNP reliabilities were higher in HOL (Mean SNP reliability: 0.42) compared to LIM (Mean SNP reliability: 0.02), underscoring the critical role of the size of the reference population in determining the accuracy and reliability of SNP effects obtained from genomic evaluations.
Conclusions
The calculation of p-values from the ssSNPBLUP framework offers an efficient approach to identify quantitative trait loci (QTL) that significantly influences traits in populations. Our approach provides a framework that could be implemented in large and complex datasets such as those used in many national routine evaluations, where only a proportion of the animals are genotyped.
