RNA-seq analysis provides a powerful tool for revealing relationships between gene expression level and biological function of proteins. In order to identify differentially expressed genes among various RNA-seq datasets obtained from different experimental designs, an appropriate normalization method for calibrating multiple experimental datasets is the first challenging problem. We propose a novel method to facilitate biologists in selecting a set of suitable housekeeping genes for inter-sample normalization. The approach is achieved by adopting user defined experimentally related keywords, GO annotations, GO term distance matrices, orthologous housekeeping gene candidates, and stability ranking of housekeeping genes. By identifying the most distanced GO terms from query keywords and selecting housekeeping gene candidates with low coefficients of variation among different spatio-temporal datasets, the proposed method can automatically enumerate a set of functionally irrelevant housekeeping genes for pratical normalization. Novel and benchmark testing RNA-seq datasets were applied to demostrate that different selections of housekeeping gene lead to strong impact on differential gene expression analysis, and compared results have shown that our proposed method outperformed other traditional approaches in terms of both sensitivity and specificity. The proposed mechanism of selecting appropriate houskeeping genes for inter-dataset normalization is robust and accurate for differential expression analyses.