Validation and phenotypic correction in genome-wide selection

The objective of this work was to evaluate the effect of the distribution of QTL effects, of the type of validation population, and of phenotype adjustment in the accuracy of genome-wide selection. Two populations of full siblings with 500 individuals were simulated, with 1,000 loci markers being genotypically considered - 100 linked to QTL. The QTL effects had uniform or exponential distribution. For validation 1, a 100-individual sample constituted the validation population; in validation 2, cross validation was applied, with a 100-individual sample in five replicates; and in validation 3, a second generation formed the validation population. The analysis methodologies used were RR-Blup and Blasso, with mixed models for phenotype correction. Without phenotypic correction, the exponential distribution led to higher accuracies, and the Blasso method showed greater accuracy with this distribution; while RR-Blup was more accurate with uniform distribution. In this scenario without correction, validations 1 and 3 were more accurate. With phenotypic correction, exponential and uniform distributions led to similar accuracies, and the Blasso method proved more accurate for both of them. In this scenario, validations 1 and 2 were more accurate. Generally, the RR-Blup method was more accurate, and the Blasso method was less biased.