Efeito da imputação de genótipos sobre a predição de característicasquantitativas de ratos utilizando marcadores genéticos

Abstract

The objective in this first chapter was to define basic concepts in quantitative and molecular genetics and to review methodologies that have been applied for genome-enabled prediction. First, models for markers effects prediction were described as Bayesian regressions (BayesA, BayesB, Bayesian LASSO, and others), semi-parametric regression (Reproducing Kernel Hilbert Spaces) and non-parametric methods (Bayesian Regularized Neural Networks). Regarding the semi and non-parametric methods, they have been proposed for genome-enabled prediction having the advantage of capturing non-linear effects between markers, which would be impossible fitting linear models. Another tool also described in this review is the genotype imputation that have been applied for fill missing data from the lab,merge data sets from different chips and even increase the number of SNPs contained in the chips.