Resumo

ALARCON-ZUNIGA, Baldomero et al. Comparison of the accuracy of genomic values of animals predicted through analysis with two alternative models. Agrociencia [online]. 2015, vol.49, n.6, pp.613-622. ISSN 2521-9766.

The genomic value is the best prediction of the genotypic value of an animal, whose accuracy varies in function of several factors. The objectives of this study were to compare, through simulation, the accuracy of the genomic values of animals predicted through analysis with two alternative models, and to obtain the genetic correlation between these and the true genetic values simulated. A population with an effective size of 800 individuals was simulated and 100 generations were used to generate linkage disequilibrium. Then, another population was simulated with 14 generations, a panel of 53 010 single nucleotide polymorphisms (SNPs), placed randomly in 30 chromosomes and 540 quantitative traits loci. The genotypes and phenotypes of 6400 animals were also simulated using a heritability=0.40, and considering only the additive effects. Four sets of molecular values were trained using the SNP markers of generations 7 to 10, and the corresponding adjusted genetic values. When the training population (TP) was larger in size, the accuracy estimators (R²) were higher. Similarly, when the evaluation population (EP) was closer to the TP, the R² estimators were higher and they became smaller when the distance between TP and EP increased. The estimate of the prediction error variance (PEV) was lower (0.15±0.01) when TP and EP were closer, regardless of the size of the TP, generation 10 for TP and generation 11 for EP. On the contrary, the highest PEV (0.29± 0.02) was obtained when the TP included animals from generation seven and generation 14 was evaluated; that is, when the distance between TP and EP was greater.

Palavras-chave : Predicted genomic value; accuracy of the genomic value; genomic selection.