Use of a Bayesian model including QTL markers increases prediction reliability when test animals are distant from the reference population.