Genome Based Search for Reference Genes for Gene Expression Analysis in Oral Cancer: a Data Science Driven Approach

Abstract

Background Quantitative real time PCR (qPCR) remains by far the most cost-effective, fast yet sensitive technique to check the gene expression levels in various systems. The traditionally used reference genes over the years were found to be regulated heavily based on sample sources and/or experimental conditions. This paper therefore presents a data science driven -omic approach for selection of reference genes from ~60,000 candidates from The Cancer Genome Atlas (TCGA) and Broad Institute Cancer Cell Line Encyclopaedia (CCLE) for gene expression studies in head and neck squamous cell carcinoma (HNSCC). mRNA-sequencing data of 500 patient samples and 33 cell lines from publicly available databases were analysed to assess stability of genes in terms of multiple statistical measures. A final set of 12 candidate genes were studied in the Indian set of data in Gene Expression Omnibus (GEO) and validated experimentally using qPCR in 35 different types of samples from platforms like drug sensitive and resistant cell lines, normal and tumor samples, fibroblast and epithelial primary culture derived from HNSCC patients from India. Result The study lead to the choice of five most stable reference genes –TYW5, RIC8B, PLEKHA3, CEP57L1 and GPR89B across three experimental platforms. Conclusion In addition to providing a set of five most stable reference genes for future gene expression analysis experiments across different types of samples in HNSCC, the study also presents an objective framework for assessing reference genes for other disease areas as well.