A prognostic 11 genes expression model for ovarian cancer.

Abstract

The symptoms of ovarian cancer at early stages are usually absent which makes the diagnosis in its early stages exceedingly difficult. Previous research has proven that ovarian cancer is a genetic disease, which depends on the alteration of multi-cancer related genes and anti-cancer genes, multi-stages and multi-pathways, involving a variety of oncogene activation and anti-oncogene inactivation. For a better understanding of the prognostic classification of ovarian cancer, gene expression profiles were used to analyze the prognostic factors of ovarian cancer, and the prognostic model was used to classify the ovarian cancer samples. The ovarian cancer samples data were downloaded from TCGA dataset. Rebust likelihood-based survival model was built to find the key genes that could function as prognostic markers. The samples were classified by unsupervised hierarchical clustering. Furthermore, Kaplan-Meier survival analysis was used to analyze the differences in the prognosis of the samples. The prognostic model was used to classify the samples, and then the best classification model was selected as the prognostic model of ovarian cancer. Finally, GEO datasets were used for external data validation. A total of 886 genes with influence on prognosis was obtained. Then genomic combinations of 11 genes were screened out by random sampling. Then the active number of influential factors was counted based on the expression level of featured genes. When the number of influencing factors is ≥7, the prognosis difference among these genes is the largest (P-value = 0.000775); and this was chosen as the final Classification model. To summary, a prognostic 11genes expression model was preliminarily built to classify the ovarian cancer samples.