Estimating accuracy of genomic breeding values for complex traits in populations without pedigree using dense markers

The aim of this study was to estimate the accuracy of genomic breeding values using GBLUP (Genomic Best Linear Unbiased Prediction) model and compare it with the traditional method with (BLUP) and without (BLUP_noPed) pedigree by computer simulation. In this study, a genome with 30 pairs of chromosomes, 3000 QTL and 45000 marker (SNP) was simulated. The different scenarios including two effective population of 100 and 1000 animals, two reference population size of 1000 and 2000 animals and three heritabilities, low (0.05), medium (0.30) and high (0.5) were investigated. The criteria were accuracy, bias and mean square error (MSE) of breeding value to compare traditional and genomic methods. The results showed that the accuracy of breeding values increased with increasing heritability. GBLUP mehtod was more accurate than BLUP_noPed in all scenarios and the former model had more performance in higher heritability compared with BLUP model. The accuracy of genomic breeding values decreased about 11% with increasing effective population size 100 to 1000; however, the accuracy of traditional methods was not affected by changing effective population size. Increasing the reference population size 1000 to 2000, the accuracy of genomic breeding values was improved by 5%. Although, the bias (underestimate) was higher for genomic than traditional methods in all scenario; but the MSE of breeding values was lower for GBLUP than BLUP model with increasing reference population size and heritability. Therefore, GBLUP method can be used to select top animals in populations without pedigree.