Derivation of Linear Models for Quantitative Traits by Bayesian Estimation with Gibbs Sampling

Abstract

Prediction of the trait values of samples based on their genetic and environmental factors is one of the vital steps in the genomic selection (GS) process used to identify valuable individuals for generating descendants in the breeding process. To obtain the estimated values of a trait, a prediction model that describes the relationship between the explanatory factors observed in the samples and the trait values is derived using a dataset consisting of factor candidates and the true values of the trait. Aside from the genetic factors obtained by the genome-wide markers, the Bayesian approach used in the derivation process is the key element for GS. Assuming a typical dataset for GS, construction of a prediction model using a linear model and determination of the model parameters using the Bayesian estimation with Gibbs sampling are explained. In addition, a sample output from the implemented software is presented.