Gene expression profiling experiments with few replications lead to great variability in the estimates of gene variances, therefore these are not robust estimates. Toward this end, several moderated t-test methods have been developed to reduce this variability and to increase power of the tests. All these methods assume a linear fixed-effects model and residual variances are smoothed under a hierarchical Bayesian framework. In this talk, we will demonstrate the implementation of these moderated t-test methods using linear mixed-effects models, where both random variances and residual variances are smoothed under the hierarchical Bayesian framework. We will apply the proposed procedure to a real gene expression data set.