Abstract
Gene expression profiling is of key importance in all domains of life sciences, as medicine, environment, and plants, for both basic and applied research. Despite the emergence of microarrays and high-throughput sequencing, qPCR remains a standard method for gene expression analyses, with its data normalization step being crucial for ensuring accuracy. Currently, the most widely used normalization method is based on the use of reference genes, assumed to be stably expressed across all experimental conditions. In the present study, we show that finding a stable combination of genes, regardless of their individual stability, outperforms standard reference genes for RT-qPCR data normalization. A stable combination of genes consists of a fixed number of genes whose individual expression balance each other all along experimental conditions of interest. Moreover, the present study shows that such an optimal combination of genes can be found using a comprehensive database of RNA-Seq data. Indeed, assuming that such a comprehensive database contains accurate gene expression profiles, we can extract in silico, by the way of the mathematical variance calculation, a stable combination of genes that reflects in vivo stability. As a case study, this new method was developed using the tomato model plant, with corresponding RNA-Seq data from the TomExpress database. However, the method is potentially applicable to other organisms with available RNA-seq data. Our results demonstrate the superiority of the reported method over commonly used housekeeping genes or other stably expressed genes. We therefore recommend the use of our new method together with classic ones in order to always obtain the best reference genes for a given experimental design.