The availability of high-density (HD) marker panels, genome wide variants and sequence data creates an unprecedented opportunity to dissect the genetic basis of complex traits, enhance genomic selection (GS) and identify causal variants of disease. The disproportional increase in the number of parameters in the genetic association model compared to the number of phenotypes has led to further deterioration in statistical power and an increase in co-linearity and false positive rates. At best, HD panels do not significantly improve GS accuracy and, at worst, reduce accuracy. This is true for both regression and variance component approaches. To remedy this situation, some form of single nucleotide polymorphisms (SNP) filtering or external information is needed. Current methods for prioritizing SNP markers (i.e. BayesB, BayesCπ) are sensitive to the increased co-linearity in HD panels which could limit their performance. In this study, the usefulness of FST, a measure of allele frequency variation among populations, as an external source of information in GS was evaluated. A simulation was carried out for a trait with heritability of 0.4. Data was divided into three subpopulations based on phenotype distribution (bottom 5%, middle 90%, top 5%). Marker data were simulated to mimic a 770 K and 1.5 million SNP marker panel. A ten-chromosome genome with 200 K and 400 K SNPs was simulated. Several scenarios with varying distributions for the quantitative trait loci (QTL) effects were simulated. Using all 200 K markers and no filtering, the accuracy of genomic prediction was 0.77. When marker effects were simulated from a gamma distribution, SNPs pre-selected based on the 99.5, 99.0 and 97.5% quantile of the FST score distribution resulted in an accuracy of 0.725, 0.797, and 0.853, respectively. Similar results were observed under other simulation scenarios. Clearly, the accuracy obtained using all SNPs can be easily achieved using only 0.5 to 1% of all markers. These results indicate that SNP filtering using already available external information could increase the accuracy of GS. This is especially important as next-generation sequencing technology becomes more affordable and accessible to human, animal and plant applications.

It becomes clear that the increase in the density of marker panels and even the use of sequence data didn’t result in any meaningful increase in the accuracy of genomic selection (GS) using either regression (RM) or variance component (VC) approaches. This is in part due to the limitations of current methods. Association model are well over-parameterized and suffer from severe co-linearity and lack of statistical power. Even when the variant effects are not directly estimated using VC based approaches, the genomic relationships didn’t improve after the marker density exceeded a certain threshold. SNP prioritization-based fixation index (FST) scores were used to track the majority of significant QTL and to reduce the dimensionality of the association model. Two populations with average LD between adjacent markers of 0.3 (P1) and 0.7 (P2) were simulated. In both populations, the genomic data consisted of 400 K SNP markers distributed on 10 chromosomes. The density of simulated genomic data mimics roughly 1.2 million SNP markers in the bovine genome. The genomic relationship matrix (G) was calculated for each set of selected SNPs based on their FST score and similar numbers of SNPs were selected randomly for comparison. Using all 400 K SNPs, 46% of the off-diagonal elements (OD) were between − 0.01 and 0.01. The same portion was 31, 23 and 16% when 80 K, 40 K and 20 K SNPs were selected based on FST scores. For randomly selected 20 K SNP subsets, around 33% of the OD fell within the same range. Genomic similarity computed using SNPs selected based on FST scores was always higher than using the same number of SNPs selected randomly. Maximum accuracies of 0.741 and 0.828 were achieved when 20 and 10 K SNPs were selected based on FST scores in P1 and P2, respectively. Genomic similarity could be maximized by the decrease in the number of selected SNPs, but it also leads to a decrease in the percentage of genetic variation explained by the selected markers. Finding the balance between these two parameters could optimize the accuracy of GS in the presence of high density marker panels.

A dramatic increase in the density of marker panels has been expected to increase the accuracy of genomic selection (GS), unfortunately, little to no improvement has been observed. By including all variants in the association model, the dimensionality of the problem should be dramatically increased, and it could undoubtedly reduce the statistical power. Using all Single nucleotide polymorphisms (SNPs) to compute the genomic relationship matrix (G) does not necessarily increase accuracy as the additive relationships can be accurately estimated using a much smaller number of markers. Due to these limitations, variant prioritization has become a necessity to improve accuracy. The fixation index (F_ST) as a measure of population differentiation has been used to identify genome segments and variants under selection pressure. Using prioritized variants has increased the accuracy of GS. Additionally, F_ST can be used to weight the relative contribution of prioritized SNPs in computing G. In this study, relative weights based on F_ST scores were developed and incorporated into the calculation of G and their impact on the estimation of variance components and accuracy was assessed. The results showed that prioritizing SNPs based on their F_ST scores resulted in an increase in the genetic similarity between training and validation animals and improved the accuracy of GS by more than 5%.

The original version of this article [1], published on 5 January 2018, contained 3 formatting errors. In this Correction the affected parts of the article are shown. The original article has been updated.