Impact of reducing the markers density on the genomic evaluation of parametric methods

The higher number of markers (p) to the number of observations (n) is the first challenge (i.e., course of dimensionality; p>>n) in genomic selection. Therefore with aim to reduce the course of dimensionality, the effect of reduce markers density on the accuracy of genomic prediction breeding values of parametric methods investigated. In this order a genome consisted of 10000 bi-allelic single nucleotide polymorphism (SNP) over 10 chromosomes, with 100 cM length each, was simulated. In this research, the different gene distributions (i.e., uniform, Gaussian and gamma), different levels of heritability (0.05, 0.25, 0.45 or 0.65) and number of quantitative trait loci (QTL; 100 or 500) were considered as assumption of simulation. Then in a selection scenario only 10% of the markers and in the other scenario 20% of the markers were selected (randomly), so that for each of the population three different marker matrices with different dimension (all, 10% and 20% of markers) were defined. According to the finding of this research, in order of preference the accuracy of genomic breeding values is influenced by the trait inheritance, the distance with the reference population, the marker density, the distribution and the number of QTLs and statistical models (p <0.001). So that in all three genetic effects distribution and with each form of the marker structure (selection or without selection of markers) it was observed that the accuracy of predictions declined significantly (p <0.05) as a result of reduction of heritability or with increasing distance from the reference population. The effect of reduce markers density on the genomic prediction accuracy of traits with different genetic architecture showed that except for traits with low heritability (0.05) and high number of QTL (500) and allele effects drown from a gamma distribution, in general reduction of marker density resulted to decrease accuracy of genomic predictions (p <0.05).