Fig. 5

The workflow of this study. (A) The gene expression profiles of 1217 breast cancer (BC) samples were obtained from the TCGA public database; (B) Four indexes, including FPKM, CV, DPM, and FC-5%, were used to select candidate RGs and Venn diagram analysis identified the RGs common to these indexes; (C) qRT-PCR experiments were carried out on various BC tissue specimens (n = 66) and BC cell lines (n = 21). P, adjacent tissues; C, cancer tissues. Six biological replicates in each tissue group and three replicates for each BC cell line were used in this study. ov, overexpression; kd, knock-down. (D) Candidate RGs were identified and evaluated by 5 public algorithms including geNorm, NormFinder, BestKeeper, ΔCt method, and ComprFinder. (E) The selected RGs were validated using 4 target genes (MAPK3, MAPK9, FAAH, and HIF1A). After being normalized by different RGs, the patterns of target gene expression were compared. The capabilities of different types of RGs were tested by correlation analysis