MODELO HIERÁRQUICO BAYESIANO NA DETERMINAÇÃO DE ASSOCIAÇÃO ENTRE MARCADORES E QTL EM UMA POPULAÇÃO F2

Abstract

The purpose of QTL mapping is to identify the position of loci controlling quantitative traits in the genome, that is, to identify the chromosomes and the locations in which the QTLs meet and, moreover, to estimate Its additive and dominance effects. To do this, a large number of molecular markers spread across the chromosomes that may be linked to these QTLs are used and, therefore, may be associated with phenotypic characteristics. Due to this, the statistical models have a high number of parameters to be estimated. However, it is expected that many of these markers will not be bound to QTL and thus some of these parameters will not be signicant. The purpose of this work is to use a priori distribution that allow the incorporation of these non-QTL-Markers associations into the model, which together with the data information lead to the updating of the QTL-binding information. Two models were used:  the first using the Lasso Bayesian shrinkage distribution (a priori) and the second Horseshoe Estimator. To verify the performance of the models, they were performed 1000 simulations referring to 10 scenarios, in which there was variation in the number of individuals, number of markers and heritability levels. It was observed that the proposed models have the ability to select the markers associated with QTL in all scenarios. The models were adjusted to grain yield data from progenies of a maize population, previously  analyzed by other methodologies, to allow comparisons of methodologies. The computational implementation of the algorithms was done using the C language and executed in the  statistical package R.