1st Leibniz MMS Days - Abstract

Reinsch, Norbert

Taking account for covariances in the Bayesian estimation of genetic marker effects in backcross-experiments

Various Bayesian approaches are in use for the estimation of genetic marker effects in planned experiments with geneticaly divergent lines or in genome wide association studies. The goal is either identification of genomic sites with impact on a trait or the prediction of genetic merit for the purpose of breeding in plants r animals. To date marker effects are routinely treated as uncorrelated. Our group has derived covariance matrices for marker effects. Four different Bayesian methods, three of them making use of such matrices, were compared in a comprehensive simulation study. Results show that taking account for covariances is beneficial both in terms of the ability to detect genomic sites with impact on the trait and with regard to the prediction ability of the model, when applied to a backcross-family. Our further research will look for generalizations to other family structures common in plant and animal breeding.