A Hybrid Method For The Implementation Of Genomic Selection Based On Fst Prioritized Single Nucleotide Polymorphisms

A Hybrid Method for the Implementation of Genomic Selection Based on Fst Prioritized Single Nucleotide Polymorphisms

Availability of high-density (HD) marker panels provides an opportunity to improve the accuracy of genomic selection (GS). Unfortunately, using HD panels resulted in no significant increase in the accuracy of GS. This lack of improvement in accuracy is more likely due to the limitations of current GS methods rather than the uselessness of HD data. Increasing variants in association models caused a reduction in statistical power. Increase in the number of genotyped animals complicated the inversion of the genomic relationship matrix. Thus, reducing the number of variants and eliminating the inversion of genomic relationship matrix are required for the full benefit from HD marker data. We proposed fixation index (FST) to prioritize SNPs for GS. To validate the usefulness of FST, a trait with heritability of 0.4 under different SNP densities was simulated. Prioritized top 2.5% markers were able to tag most significant QTL and to increase functional genomic similarity. The latter could be used as a decision-making or selection tool. In spite of being able to prioritize markers in linkage disequilibrium with relevant QTL, the latter explained only a portion of the genetic variance. This is the case because small effects QTL are often not tagged with the prioritized SNPs. These small effect QTL could be tracked, however, by a polygenic component. Thus, a hybrid model was proposed that included the prioritized SNPs and a polygenic component in the association model. The proposed approach was evaluated based on simulated data of a trait with heritability of 0.1 and 0.4 and a real data of weaning weight in beef cattle. Using only genotyped animals, the hybrid model outperformed BayesB, BayesC and GBLUP when the prioritized 2.5% SNPs were used in the association model. The hybrid model was extended to accommodate non-genotyped animals. It outperformed ssGBLUP method using simulated data under both heritability scenarios. Although the results of the evaluation are likely to depend on the data generating process including the genetic complexity of the trait, the hybrid model seemed to be competitive compared to current methods. Furthermore, its computational costs in terms of CPU time and peak memory are limited.

Application of genomics in livestock populations under selection or conservation

Index Medicus

Vols. for 1963- include as pt. 2 of the Jan. issue: Medical subject headings.