Modelo hierárquico generalizado normal assimétrico Bayesiano aplicado à análise genômica

Abstract

A crucial point in genomic analysis is the correct selection of genetically superior individuals for characters of economic importance. In this dissertation we study the application of the Generalized Hierarchical Bayesian Model (MHGB) using the asymetric gaussian distribution to the Genome wide Selection problem (GWS). The reasoning for this choice of modelling is to challenge current models of GWS when they fail their assumptions and become less reliable. A simulation study was carried to compare reference models to MHGB. Markers of actual SNPs data where used to simulated phenotypes in different scenarios for number of genes and heritability, as well as degrees of asymmetry in the error distribution. In symmetric scenarios MHGB was almost as acurate as main reference methods GBLUP. When asymetry arises, MHGB accuracy overtakes GBLUP and all other considered methods. There are evidence that MHGB should be used with advantages in GWS, whenever asymmetries are identified in the data distribution.