The use of mixed models to identify differentially expressed genes when a single replicate per biological condition is present

In microarray data analysis, selection of biologically relevant genes are essential tasks. In this study, we propose a novel statistical procedure based on a linear mixed-effects model to identify the differentially expressed genes in a preprocessed cDNA microarray experiments. This method uses standardized conditional residuals to perform gene-specific comparisons of experimental conditions. The novelty of this approach is that it enables hypothesis testing when only a single replicate per experimental condition is present. This method accommodates a wide variety of experimental designs and can simultaneously assess differences between multiple types of biological samples. Interestingly, the method can be applied for cDNA as well as oligonucleotide microarray experiments. The method is illustrated using two publicly available gene expression datasets. Simulations show that the tests developed here control the Type-I error and have enough power to detect biologically important genes