The comparison of the accuracy of genomic predictions for low heritability traits under single-step and multi-step GBLUP methods applying simulated data

Reproductive traits are considered to be as functional traits in the dairy breeding industry due to their great impact on the profitability. Due to low heritability of reproductive traits, genetic progress in these traits is slow. Therefore, genomic selection may be considered as an effective method for increasing genetic improvement for these traits. The accuracy of genomic selection depends on the several factors such as heritability of traits, statistical method for calculating SNP marker effects in reference population and number of individuals in the reference population. The present study was aimed to compare the accuracy and bias of genomic predictions for low heritability traits using SS-GBLUP and MS-GBLUP methods under different strategies applying simulated data. For this purpose, a genome containing 1000, 2000 and 5000 single-nucleotide polymorphism with two alleles (SNP) with length of one cM was simulated. The numbers of individuals in reference population were considered to be 1000, 1500 and 2000, respectively. The mean accuracies were estimated as 0.244 and 0.399 under above methods, respectively. Also, with an increase in the numbers of individual in the reference population, from 1000 to 2000 the mean accuracy of genomic prediction under above methods were increased from 0.209 to 0.294 and from 0.348 to 0.460, respectively. The averages for regression coefficient under above methods were estimated to be 1.2 and 0.94, respectively. Our results indicated that SS-GBLUP method along with an increase in the numbers of individuals' reference population, is suggested to increase the accuracy of selection for low-heritability reproduction traits.