Identifying differentially expressed (DE) genes between cancer and normal tissues is of basic importance for studying cancer mechanisms. However, current methods, such as the commonly used Significance Analysis of Microarrays (SAM), are biased to genes with low expression levels. Recently, we proposed an algorithm, named the pairwise difference (PD) algorithm, to identify highly expressed DE genes based on reproducibility evaluation of top-ranked expression differences between paired technical replicates of cells under two experimental conditions. In this study, we extended the application of the algorithm to the identification of DE genes between two types of tissue samples (biological replicates) based on several independent datasets or sub-datasets of a dataset, by constructing multiple paired average gene expression profiles for the two types of samples. Using multiple datasets for lung and esophageal cancers, we demonstrated that PD could identify many DE genes highly expressed in both cancer and normal tissues that tended to be missed by the commonly used SAM. These highly expressed DE genes, including many housekeeping genes, were significantly enriched in many conservative pathways, such as ribosome, proteasome, phagosome and TNF signaling pathways with important functional significances in oncogenesis.