Gravar-mail: Applying dynamic Bayesian networks to perturbed gene expression data