Comparing Gene Expression Data From Formalin-Fixed, Paraffin Embedded Tissues and qPCR With That From Snap-Frozen Tissue and Microarrays for Modeling Outcomes of Patients With Ovarian Carcinoma

open access print version - Medscape (Wisconsin) 
 Comparing Gene Expression Data From Formalin-Fixed, Paraffin Embedded Tissues and qPCR With That From Snap-Frozen Tissue and Microarrays for Modeling Outcomes of Patients With Ovarian Carcinoma

 Background
Using gene expression and clinical data from The Cancer Genome Atlas (TCGA), we previously developed a model that predicts variations in response of high-grade serous ovarian cancer to cytotoxic chemotherapies. In that publication,^[1] we described a method for reducing the list of genes needed to predict clinical outcomes to fewer than 100. We selected those genes from more than 10,000 possibilities by identifying genes within a core group of 12 cancer pathways^[2,3] whose variation in expression had the greatest effect on disease progression. Predictions of response to specific chemotherapeutic agents were suggested by the cumulative levels of gene expression among the 91 genes selected from the 12 pathways. Three of the pathways did not have genes identified, leaving 9 core pathways informative. We defined the predictions made by gene expression within these pathways as the Patient-Specific Risk Profile (PSRP)......

Gene List
The genes whose expression we assayed were selected from a gene set constituting the 9 core pathways described previously.^[1] 91 genes were chosen from our previously published PSRP results. Analysis was performed according to the 9 core pathways as well as a revised six-gene set representing the neurotrophin pathway, making a total of 10 pathways available for analysis (Additional file 1: Table S1 http://www.biomedcentral.com/1472-6890/15/17/additional). The subsets of genes used to define a pathway's expression are listed in the Supplement as well (Additional file 1: Table S2 http://www.biomedcentral.com/1472-6890/15/17/additional). We used the housekeeping genes glyceraldehyde 3-phosphate dehydrogenase (GAPDH), hypoxanthine phosphoribosyltransferase 1 (HPRT1), and beta-D-glucuronidase (GUSB) to normalize gene expression.

In summary, this study validated the use of FFPE tissue and qPCR—instead of snap-frozen tissue and microarrays—to obtain gene expression data for core cellular pathways. This supports the use these tissue samples when predictive modeling of ovarian cancer was done in larger data sets such as the TCGA. The variation in expression noted between our different samples does not appear to significantly distort expected outputs, leading us to believe that a model derived from expression reported using one approach could be used with a more convenient and "real world" approach when evaluating clinical samples. Our assay will be tested in a recurrent disease setting to more definitively evaluate predictive capacity prospectively.