Comparison of GWS Methods for the Selection of Higher Individuals

The aim of this study was to compare rrBLUP, gBLUP and LASSO methods used in genome-wide selection, regarding the ability of selecting genetically superior individuals. F2 populations were simulated with different sizes (100, 500, and 1000 individuals) with 10 replication each consisting of 10 GL with 100cM each containing 100 equally spaced markers in linkage group with a total of 1,000 molecular markers, 200 that controlled the traits defined as 20 those for each initial GL. Genetic and phenotypic values with binomial distribution of effects for each GL, and lack of dominance to phenotypic values considered heritability of 20, 50 and 80% were generated. Analyses were performed in software GENES and R. For comparison of methodologies was obtained processing time of the analysis, the coefficient of coincidence (selection of 5, 10 and 20% of the best individuals) and Spearman correlation between true breeding values, genomic and predicted values for each methodology. Considering the processing time of the three methods were statistically different, where rrBLUP was the fastest and the slowest was LASSO analysis. In Spearman correlation was concluded that the rrBLUP and GBLUP are statistically equal and the LASSO which is provided under correlation values. A similar result was obtained in the match variables between selected individuals, in which the LASSO is statistically different having less than other methodologies value. Therefore, for the scenarios evaluated is concluded that the LASSO is the worst method of selection of genetically superior individuals.