Application of Bayesian Elastic Net and Other Shrinkage Methods in Genomic Selection and QTL Mapping

摘要

The Elastic Net is a variable selection and shrinkage estimation method especially designed for regression settings with a large number of correlated predictors. Recently, a Bayesian formulation of the Elastic Net was proposed (BEN=Bayesian Elastic Net). In this article, we extend the BEN to model the combined effects of dense molecular markers and pedigree data and evaluate the performance of the proposed model using a barley data set and two large wheat data sets. The predictive power of the proposed model was compared with those of two well-established models: the Bayesian LASSO and the Bayesian Ridge Regression. Results show that the prediction assessment of BEN was as accurate as those of the other methods in all studied cases. The number of molecular markers with significant effects detected by BEN in four data sets was compared with those found by the Bayesian LASSO and Bayesian Ridge Regression models. An R-program that implements the proposed model is available.