Bimodally Expressed Genes in Ovarian Carcinoma.
Allison HS Hall, Dawn N Kernagis, Michael B Datto. Duke University Medical Center, Durham, NC
Background: In certain cancers, genes can behave as molecular switches, having either an "on" or "off" state. These genes generally have what is statistically described as a bimodal distribution in expression. In the context of breast cancer, these switches are well described. They include ER and HER2 and define distinct cancer subtypes, which respond to different treatments and have different overall survival. Genes with a bimodal pattern of expression are also good targets for clinical testing because the difference in expression between "on" versus "off" populations is easy to measure. We hypothesized that bimodally expressed genes can be found in other cancers, specifically ovarian cancer, and that in this context they also define important biological tumor subtypes.
Design: We evaluated the largest publically available ovarian cancer dataset, which includes 245 malignant serous tumors, 18 serous tumors of low malignant potential and 20 malignant endometrioid tumors. We used a bimodal gene discovery algorithm to identify genes with a bimodal distribution. Correlations among expression of the bimodal genes, overall survival and tumor type were investigated using Fisher's exact test and standard Kaplan Meier survival analysis.
Results: We found many genes with prominent bimodal distributions of expression. Among the 174 genes with the strongest bimodal expression, we found ten genes that had a significant association with overall survival in malignant serous tumors, including placental alkaline phosphatase, JUB and interferon alpha 2. When expression of these genes was combined into a single survival index, the median survival for patients with malignant serous tumors with a favorable survival score was 65 months, versus 33 months for patients with malignant serous tumors with an unfavorable survival score (p<0.0001). We also found thirteen genes with significantly different expression between malignant serous tumors and endometrioid tumors and twenty-six genes with significantly different expression between malignant serous tumors and serous tumors of low malignant potential.
Conclusions: This work demonstrates that ovarian cancer, like breast cancer, can be classified by bimodal gene expression patterns. This approach provides insight into clinical behavior and identifies ideal targets for robust and precise clinical testing.
Category: Special Category - Pan-genomic/Pan-proteomic approaches to Cancer
Monday, February 28, 2011 9:30 AM
Poster Session I Stowell-Orbison/Surgical Pathology/Autopsy Awards Poster Session # 243, Monday Morning